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"
44 #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 * @node: hash table node
79 * @rcu: rcu struct for tipc_sock
90 struct list_head sock_list;
91 struct list_head publications;
94 unsigned long probing_intv;
100 struct rhash_head node;
104 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb);
105 static void tipc_data_ready(struct sock *sk);
106 static void tipc_write_space(struct sock *sk);
107 static int tipc_release(struct socket *sock);
108 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags);
109 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p);
110 static void tipc_sk_timeout(unsigned long data);
111 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
112 struct tipc_name_seq const *seq);
113 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
114 struct tipc_name_seq const *seq);
115 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
116 static int tipc_sk_insert(struct tipc_sock *tsk);
117 static void tipc_sk_remove(struct tipc_sock *tsk);
119 static const struct proto_ops packet_ops;
120 static const struct proto_ops stream_ops;
121 static const struct proto_ops msg_ops;
123 static struct proto tipc_proto;
124 static struct proto tipc_proto_kern;
126 static const struct nla_policy tipc_nl_sock_policy[TIPC_NLA_SOCK_MAX + 1] = {
127 [TIPC_NLA_SOCK_UNSPEC] = { .type = NLA_UNSPEC },
128 [TIPC_NLA_SOCK_ADDR] = { .type = NLA_U32 },
129 [TIPC_NLA_SOCK_REF] = { .type = NLA_U32 },
130 [TIPC_NLA_SOCK_CON] = { .type = NLA_NESTED },
131 [TIPC_NLA_SOCK_HAS_PUBL] = { .type = NLA_FLAG }
135 * Revised TIPC socket locking policy:
137 * Most socket operations take the standard socket lock when they start
138 * and hold it until they finish (or until they need to sleep). Acquiring
139 * this lock grants the owner exclusive access to the fields of the socket
140 * data structures, with the exception of the backlog queue. A few socket
141 * operations can be done without taking the socket lock because they only
142 * read socket information that never changes during the life of the socket.
144 * Socket operations may acquire the lock for the associated TIPC port if they
145 * need to perform an operation on the port. If any routine needs to acquire
146 * both the socket lock and the port lock it must take the socket lock first
147 * to avoid the risk of deadlock.
149 * The dispatcher handling incoming messages cannot grab the socket lock in
150 * the standard fashion, since invoked it runs at the BH level and cannot block.
151 * Instead, it checks to see if the socket lock is currently owned by someone,
152 * and either handles the message itself or adds it to the socket's backlog
153 * queue; in the latter case the queued message is processed once the process
154 * owning the socket lock releases it.
156 * NOTE: Releasing the socket lock while an operation is sleeping overcomes
157 * the problem of a blocked socket operation preventing any other operations
158 * from occurring. However, applications must be careful if they have
159 * multiple threads trying to send (or receive) on the same socket, as these
160 * operations might interfere with each other. For example, doing a connect
161 * and a receive at the same time might allow the receive to consume the
162 * ACK message meant for the connect. While additional work could be done
163 * to try and overcome this, it doesn't seem to be worthwhile at the present.
165 * NOTE: Releasing the socket lock while an operation is sleeping also ensures
166 * that another operation that must be performed in a non-blocking manner is
167 * not delayed for very long because the lock has already been taken.
169 * NOTE: This code assumes that certain fields of a port/socket pair are
170 * constant over its lifetime; such fields can be examined without taking
171 * the socket lock and/or port lock, and do not need to be re-read even
172 * after resuming processing after waiting. These fields include:
174 * - pointer to socket sk structure (aka tipc_sock structure)
175 * - pointer to port structure
179 static u32 tsk_own_node(struct tipc_sock *tsk)
181 return msg_prevnode(&tsk->phdr);
184 static u32 tsk_peer_node(struct tipc_sock *tsk)
186 return msg_destnode(&tsk->phdr);
189 static u32 tsk_peer_port(struct tipc_sock *tsk)
191 return msg_destport(&tsk->phdr);
194 static bool tsk_unreliable(struct tipc_sock *tsk)
196 return msg_src_droppable(&tsk->phdr) != 0;
199 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
201 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
204 static bool tsk_unreturnable(struct tipc_sock *tsk)
206 return msg_dest_droppable(&tsk->phdr) != 0;
209 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
211 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
214 static int tsk_importance(struct tipc_sock *tsk)
216 return msg_importance(&tsk->phdr);
219 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
221 if (imp > TIPC_CRITICAL_IMPORTANCE)
223 msg_set_importance(&tsk->phdr, (u32)imp);
227 static struct tipc_sock *tipc_sk(const struct sock *sk)
229 return container_of(sk, struct tipc_sock, sk);
232 static int tsk_conn_cong(struct tipc_sock *tsk)
234 return tsk->sent_unacked >= TIPC_FLOWCTRL_WIN;
238 * tsk_advance_rx_queue - discard first buffer in socket receive queue
240 * Caller must hold socket lock
242 static void tsk_advance_rx_queue(struct sock *sk)
244 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
248 * tsk_rej_rx_queue - reject all buffers in socket receive queue
250 * Caller must hold socket lock
252 static void tsk_rej_rx_queue(struct sock *sk)
256 u32 own_node = tsk_own_node(tipc_sk(sk));
258 while ((skb = __skb_dequeue(&sk->sk_receive_queue))) {
259 if (tipc_msg_reverse(own_node, skb, &dnode, TIPC_ERR_NO_PORT))
260 tipc_link_xmit_skb(sock_net(sk), skb, dnode, 0);
264 /* tsk_peer_msg - verify if message was sent by connected port's peer
266 * Handles cases where the node's network address has changed from
267 * the default of <0.0.0> to its configured setting.
269 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
271 struct tipc_net *tn = net_generic(sock_net(&tsk->sk), tipc_net_id);
272 u32 peer_port = tsk_peer_port(tsk);
276 if (unlikely(!tsk->connected))
279 if (unlikely(msg_origport(msg) != peer_port))
282 orig_node = msg_orignode(msg);
283 peer_node = tsk_peer_node(tsk);
285 if (likely(orig_node == peer_node))
288 if (!orig_node && (peer_node == tn->own_addr))
291 if (!peer_node && (orig_node == tn->own_addr))
298 * tipc_sk_create - create a TIPC socket
299 * @net: network namespace (must be default network)
300 * @sock: pre-allocated socket structure
301 * @protocol: protocol indicator (must be 0)
302 * @kern: caused by kernel or by userspace?
304 * This routine creates additional data structures used by the TIPC socket,
305 * initializes them, and links them together.
307 * Returns 0 on success, errno otherwise
309 static int tipc_sk_create(struct net *net, struct socket *sock,
310 int protocol, int kern)
313 const struct proto_ops *ops;
316 struct tipc_sock *tsk;
317 struct tipc_msg *msg;
319 /* Validate arguments */
320 if (unlikely(protocol != 0))
321 return -EPROTONOSUPPORT;
323 switch (sock->type) {
326 state = SS_UNCONNECTED;
330 state = SS_UNCONNECTED;
341 /* Allocate socket's protocol area */
343 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto);
345 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto_kern);
351 tsk->max_pkt = MAX_PKT_DEFAULT;
352 INIT_LIST_HEAD(&tsk->publications);
354 tn = net_generic(sock_net(sk), tipc_net_id);
355 tipc_msg_init(tn->own_addr, msg, TIPC_LOW_IMPORTANCE, TIPC_NAMED_MSG,
358 /* Finish initializing socket data structures */
361 sock_init_data(sock, sk);
362 if (tipc_sk_insert(tsk)) {
363 pr_warn("Socket create failed; port numbrer exhausted\n");
366 msg_set_origport(msg, tsk->portid);
367 setup_timer(&sk->sk_timer, tipc_sk_timeout, (unsigned long)tsk);
368 sk->sk_backlog_rcv = tipc_backlog_rcv;
369 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
370 sk->sk_data_ready = tipc_data_ready;
371 sk->sk_write_space = tipc_write_space;
372 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
373 tsk->sent_unacked = 0;
374 atomic_set(&tsk->dupl_rcvcnt, 0);
376 if (sock->state == SS_READY) {
377 tsk_set_unreturnable(tsk, true);
378 if (sock->type == SOCK_DGRAM)
379 tsk_set_unreliable(tsk, true);
385 * tipc_sock_create_local - create TIPC socket from inside TIPC module
386 * @type: socket type - SOCK_RDM or SOCK_SEQPACKET
388 * We cannot use sock_creat_kern here because it bumps module user count.
389 * Since socket owner and creator is the same module we must make sure
390 * that module count remains zero for module local sockets, otherwise
391 * we cannot do rmmod.
393 * Returns 0 on success, errno otherwise
395 int tipc_sock_create_local(struct net *net, int type, struct socket **res)
399 rc = sock_create_lite(AF_TIPC, type, 0, res);
401 pr_err("Failed to create kernel socket\n");
404 tipc_sk_create(net, *res, 0, 1);
410 * tipc_sock_release_local - release socket created by tipc_sock_create_local
411 * @sock: the socket to be released.
413 * Module reference count is not incremented when such sockets are created,
414 * so we must keep it from being decremented when they are released.
416 void tipc_sock_release_local(struct socket *sock)
424 * tipc_sock_accept_local - accept a connection on a socket created
425 * with tipc_sock_create_local. Use this function to avoid that
426 * module reference count is inadvertently incremented.
428 * @sock: the accepting socket
429 * @newsock: reference to the new socket to be created
430 * @flags: socket flags
433 int tipc_sock_accept_local(struct socket *sock, struct socket **newsock,
436 struct sock *sk = sock->sk;
439 ret = sock_create_lite(sk->sk_family, sk->sk_type,
440 sk->sk_protocol, newsock);
444 ret = tipc_accept(sock, *newsock, flags);
446 sock_release(*newsock);
449 (*newsock)->ops = sock->ops;
453 static void tipc_sk_callback(struct rcu_head *head)
455 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
461 * tipc_release - destroy a TIPC socket
462 * @sock: socket to destroy
464 * This routine cleans up any messages that are still queued on the socket.
465 * For DGRAM and RDM socket types, all queued messages are rejected.
466 * For SEQPACKET and STREAM socket types, the first message is rejected
467 * and any others are discarded. (If the first message on a STREAM socket
468 * is partially-read, it is discarded and the next one is rejected instead.)
470 * NOTE: Rejected messages are not necessarily returned to the sender! They
471 * are returned or discarded according to the "destination droppable" setting
472 * specified for the message by the sender.
474 * Returns 0 on success, errno otherwise
476 static int tipc_release(struct socket *sock)
478 struct sock *sk = sock->sk;
480 struct tipc_sock *tsk;
482 u32 dnode, probing_state;
485 * Exit if socket isn't fully initialized (occurs when a failed accept()
486 * releases a pre-allocated child socket that was never used)
496 * Reject all unreceived messages, except on an active connection
497 * (which disconnects locally & sends a 'FIN+' to peer)
499 dnode = tsk_peer_node(tsk);
500 while (sock->state != SS_DISCONNECTING) {
501 skb = __skb_dequeue(&sk->sk_receive_queue);
504 if (TIPC_SKB_CB(skb)->handle != NULL)
507 if ((sock->state == SS_CONNECTING) ||
508 (sock->state == SS_CONNECTED)) {
509 sock->state = SS_DISCONNECTING;
511 tipc_node_remove_conn(net, dnode, tsk->portid);
513 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
515 tipc_link_xmit_skb(net, skb, dnode, 0);
519 tipc_sk_withdraw(tsk, 0, NULL);
520 probing_state = tsk->probing_state;
521 if (del_timer_sync(&sk->sk_timer) &&
522 probing_state != TIPC_CONN_PROBING)
525 if (tsk->connected) {
526 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
527 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
528 tsk_own_node(tsk), tsk_peer_port(tsk),
529 tsk->portid, TIPC_ERR_NO_PORT);
531 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
532 tipc_node_remove_conn(net, dnode, tsk->portid);
535 /* Discard any remaining (connection-based) messages in receive queue */
536 __skb_queue_purge(&sk->sk_receive_queue);
538 /* Reject any messages that accumulated in backlog queue */
539 sock->state = SS_DISCONNECTING;
542 call_rcu(&tsk->rcu, tipc_sk_callback);
549 * tipc_bind - associate or disassocate TIPC name(s) with a socket
550 * @sock: socket structure
551 * @uaddr: socket address describing name(s) and desired operation
552 * @uaddr_len: size of socket address data structure
554 * Name and name sequence binding is indicated using a positive scope value;
555 * a negative scope value unbinds the specified name. Specifying no name
556 * (i.e. a socket address length of 0) unbinds all names from the socket.
558 * Returns 0 on success, errno otherwise
560 * NOTE: This routine doesn't need to take the socket lock since it doesn't
561 * access any non-constant socket information.
563 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
566 struct sock *sk = sock->sk;
567 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
568 struct tipc_sock *tsk = tipc_sk(sk);
572 if (unlikely(!uaddr_len)) {
573 res = tipc_sk_withdraw(tsk, 0, NULL);
577 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
581 if (addr->family != AF_TIPC) {
586 if (addr->addrtype == TIPC_ADDR_NAME)
587 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
588 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
593 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
594 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
595 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
600 res = (addr->scope > 0) ?
601 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
602 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
609 * tipc_getname - get port ID of socket or peer socket
610 * @sock: socket structure
611 * @uaddr: area for returned socket address
612 * @uaddr_len: area for returned length of socket address
613 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
615 * Returns 0 on success, errno otherwise
617 * NOTE: This routine doesn't need to take the socket lock since it only
618 * accesses socket information that is unchanging (or which changes in
619 * a completely predictable manner).
621 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
622 int *uaddr_len, int peer)
624 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
625 struct tipc_sock *tsk = tipc_sk(sock->sk);
626 struct tipc_net *tn = net_generic(sock_net(sock->sk), tipc_net_id);
628 memset(addr, 0, sizeof(*addr));
630 if ((sock->state != SS_CONNECTED) &&
631 ((peer != 2) || (sock->state != SS_DISCONNECTING)))
633 addr->addr.id.ref = tsk_peer_port(tsk);
634 addr->addr.id.node = tsk_peer_node(tsk);
636 addr->addr.id.ref = tsk->portid;
637 addr->addr.id.node = tn->own_addr;
640 *uaddr_len = sizeof(*addr);
641 addr->addrtype = TIPC_ADDR_ID;
642 addr->family = AF_TIPC;
644 addr->addr.name.domain = 0;
650 * tipc_poll - read and possibly block on pollmask
651 * @file: file structure associated with the socket
652 * @sock: socket for which to calculate the poll bits
655 * Returns pollmask value
658 * It appears that the usual socket locking mechanisms are not useful here
659 * since the pollmask info is potentially out-of-date the moment this routine
660 * exits. TCP and other protocols seem to rely on higher level poll routines
661 * to handle any preventable race conditions, so TIPC will do the same ...
663 * TIPC sets the returned events as follows:
665 * socket state flags set
666 * ------------ ---------
667 * unconnected no read flags
668 * POLLOUT if port is not congested
670 * connecting POLLIN/POLLRDNORM if ACK/NACK in rx queue
673 * connected POLLIN/POLLRDNORM if data in rx queue
674 * POLLOUT if port is not congested
676 * disconnecting POLLIN/POLLRDNORM/POLLHUP
679 * listening POLLIN if SYN in rx queue
682 * ready POLLIN/POLLRDNORM if data in rx queue
683 * [connectionless] POLLOUT (since port cannot be congested)
685 * IMPORTANT: The fact that a read or write operation is indicated does NOT
686 * imply that the operation will succeed, merely that it should be performed
687 * and will not block.
689 static unsigned int tipc_poll(struct file *file, struct socket *sock,
692 struct sock *sk = sock->sk;
693 struct tipc_sock *tsk = tipc_sk(sk);
696 sock_poll_wait(file, sk_sleep(sk), wait);
698 switch ((int)sock->state) {
705 if (!tsk->link_cong && !tsk_conn_cong(tsk))
710 if (!skb_queue_empty(&sk->sk_receive_queue))
711 mask |= (POLLIN | POLLRDNORM);
713 case SS_DISCONNECTING:
714 mask = (POLLIN | POLLRDNORM | POLLHUP);
722 * tipc_sendmcast - send multicast message
723 * @sock: socket structure
724 * @seq: destination address
725 * @msg: message to send
726 * @dsz: total length of message data
727 * @timeo: timeout to wait for wakeup
729 * Called from function tipc_sendmsg(), which has done all sanity checks
730 * Returns the number of bytes sent on success, or errno
732 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
733 struct msghdr *msg, size_t dsz, long timeo)
735 struct sock *sk = sock->sk;
736 struct tipc_sock *tsk = tipc_sk(sk);
737 struct net *net = sock_net(sk);
738 struct tipc_msg *mhdr = &tsk->phdr;
739 struct sk_buff_head *pktchain = &sk->sk_write_queue;
740 struct iov_iter save = msg->msg_iter;
744 msg_set_type(mhdr, TIPC_MCAST_MSG);
745 msg_set_lookup_scope(mhdr, TIPC_CLUSTER_SCOPE);
746 msg_set_destport(mhdr, 0);
747 msg_set_destnode(mhdr, 0);
748 msg_set_nametype(mhdr, seq->type);
749 msg_set_namelower(mhdr, seq->lower);
750 msg_set_nameupper(mhdr, seq->upper);
751 msg_set_hdr_sz(mhdr, MCAST_H_SIZE);
754 mtu = tipc_bclink_get_mtu();
755 rc = tipc_msg_build(mhdr, msg, 0, dsz, mtu, pktchain);
756 if (unlikely(rc < 0))
760 rc = tipc_bclink_xmit(net, pktchain);
761 if (likely(rc >= 0)) {
765 if (rc == -EMSGSIZE) {
766 msg->msg_iter = save;
769 if (rc != -ELINKCONG)
771 tipc_sk(sk)->link_cong = 1;
772 rc = tipc_wait_for_sndmsg(sock, &timeo);
774 __skb_queue_purge(pktchain);
780 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
781 * @arrvq: queue with arriving messages, to be cloned after destination lookup
782 * @inputq: queue with cloned messages, delivered to socket after dest lookup
784 * Multi-threaded: parallel calls with reference to same queues may occur
786 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
787 struct sk_buff_head *inputq)
789 struct tipc_msg *msg;
790 struct tipc_plist dports;
792 u32 scope = TIPC_CLUSTER_SCOPE;
793 struct sk_buff_head tmpq;
795 struct sk_buff *skb, *_skb;
797 __skb_queue_head_init(&tmpq);
798 tipc_plist_init(&dports);
800 skb = tipc_skb_peek(arrvq, &inputq->lock);
801 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
803 hsz = skb_headroom(skb) + msg_hdr_sz(msg);
805 if (in_own_node(net, msg_orignode(msg)))
806 scope = TIPC_NODE_SCOPE;
808 /* Create destination port list and message clones: */
809 tipc_nametbl_mc_translate(net,
810 msg_nametype(msg), msg_namelower(msg),
811 msg_nameupper(msg), scope, &dports);
812 portid = tipc_plist_pop(&dports);
813 for (; portid; portid = tipc_plist_pop(&dports)) {
814 _skb = __pskb_copy(skb, hsz, GFP_ATOMIC);
816 msg_set_destport(buf_msg(_skb), portid);
817 __skb_queue_tail(&tmpq, _skb);
820 pr_warn("Failed to clone mcast rcv buffer\n");
822 /* Append to inputq if not already done by other thread */
823 spin_lock_bh(&inputq->lock);
824 if (skb_peek(arrvq) == skb) {
825 skb_queue_splice_tail_init(&tmpq, inputq);
826 kfree_skb(__skb_dequeue(arrvq));
828 spin_unlock_bh(&inputq->lock);
829 __skb_queue_purge(&tmpq);
832 tipc_sk_rcv(net, inputq);
836 * tipc_sk_proto_rcv - receive a connection mng protocol message
837 * @tsk: receiving socket
838 * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
840 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
842 struct tipc_msg *msg = buf_msg(*skb);
845 u32 own_node = tsk_own_node(tsk);
846 /* Ignore if connection cannot be validated: */
847 if (!tsk_peer_msg(tsk, msg))
850 tsk->probing_state = TIPC_CONN_OK;
852 if (msg_type(msg) == CONN_ACK) {
853 conn_cong = tsk_conn_cong(tsk);
854 tsk->sent_unacked -= msg_msgcnt(msg);
856 tsk->sk.sk_write_space(&tsk->sk);
857 } else if (msg_type(msg) == CONN_PROBE) {
858 if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
859 msg_set_type(msg, CONN_PROBE_REPLY);
863 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
869 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
871 struct sock *sk = sock->sk;
872 struct tipc_sock *tsk = tipc_sk(sk);
877 int err = sock_error(sk);
880 if (sock->state == SS_DISCONNECTING)
884 if (signal_pending(current))
885 return sock_intr_errno(*timeo_p);
887 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
888 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
889 finish_wait(sk_sleep(sk), &wait);
895 * tipc_sendmsg - send message in connectionless manner
896 * @iocb: if NULL, indicates that socket lock is already held
897 * @sock: socket structure
898 * @m: message to send
899 * @dsz: amount of user data to be sent
901 * Message must have an destination specified explicitly.
902 * Used for SOCK_RDM and SOCK_DGRAM messages,
903 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
904 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
906 * Returns the number of bytes sent on success, or errno otherwise
908 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
909 struct msghdr *m, size_t dsz)
911 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
912 struct sock *sk = sock->sk;
913 struct tipc_sock *tsk = tipc_sk(sk);
914 struct net *net = sock_net(sk);
915 struct tipc_msg *mhdr = &tsk->phdr;
917 struct sk_buff_head *pktchain = &sk->sk_write_queue;
919 struct tipc_name_seq *seq = &dest->addr.nameseq;
920 struct iov_iter save;
926 return -EDESTADDRREQ;
928 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
929 (dest->family != AF_TIPC)))
932 if (dsz > TIPC_MAX_USER_MSG_SIZE)
938 if (unlikely(sock->state != SS_READY)) {
939 if (sock->state == SS_LISTENING) {
943 if (sock->state != SS_UNCONNECTED) {
947 if (tsk->published) {
951 if (dest->addrtype == TIPC_ADDR_NAME) {
952 tsk->conn_type = dest->addr.name.name.type;
953 tsk->conn_instance = dest->addr.name.name.instance;
957 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
959 if (dest->addrtype == TIPC_ADDR_MCAST) {
960 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
962 } else if (dest->addrtype == TIPC_ADDR_NAME) {
963 u32 type = dest->addr.name.name.type;
964 u32 inst = dest->addr.name.name.instance;
965 u32 domain = dest->addr.name.domain;
968 msg_set_type(mhdr, TIPC_NAMED_MSG);
969 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
970 msg_set_nametype(mhdr, type);
971 msg_set_nameinst(mhdr, inst);
972 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
973 dport = tipc_nametbl_translate(net, type, inst, &dnode);
974 msg_set_destnode(mhdr, dnode);
975 msg_set_destport(mhdr, dport);
976 if (unlikely(!dport && !dnode)) {
980 } else if (dest->addrtype == TIPC_ADDR_ID) {
981 dnode = dest->addr.id.node;
982 msg_set_type(mhdr, TIPC_DIRECT_MSG);
983 msg_set_lookup_scope(mhdr, 0);
984 msg_set_destnode(mhdr, dnode);
985 msg_set_destport(mhdr, dest->addr.id.ref);
986 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
991 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
992 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
997 skb = skb_peek(pktchain);
998 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
999 rc = tipc_link_xmit(net, pktchain, dnode, tsk->portid);
1000 if (likely(rc >= 0)) {
1001 if (sock->state != SS_READY)
1002 sock->state = SS_CONNECTING;
1006 if (rc == -EMSGSIZE) {
1010 if (rc != -ELINKCONG)
1013 rc = tipc_wait_for_sndmsg(sock, &timeo);
1015 __skb_queue_purge(pktchain);
1024 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1026 struct sock *sk = sock->sk;
1027 struct tipc_sock *tsk = tipc_sk(sk);
1032 int err = sock_error(sk);
1035 if (sock->state == SS_DISCONNECTING)
1037 else if (sock->state != SS_CONNECTED)
1041 if (signal_pending(current))
1042 return sock_intr_errno(*timeo_p);
1044 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1045 done = sk_wait_event(sk, timeo_p,
1047 !tsk_conn_cong(tsk)) ||
1049 finish_wait(sk_sleep(sk), &wait);
1055 * tipc_send_stream - send stream-oriented data
1057 * @sock: socket structure
1059 * @dsz: total length of data to be transmitted
1061 * Used for SOCK_STREAM data.
1063 * Returns the number of bytes sent on success (or partial success),
1064 * or errno if no data sent
1066 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1067 struct msghdr *m, size_t dsz)
1069 struct sock *sk = sock->sk;
1070 struct net *net = sock_net(sk);
1071 struct tipc_sock *tsk = tipc_sk(sk);
1072 struct tipc_msg *mhdr = &tsk->phdr;
1073 struct sk_buff_head *pktchain = &sk->sk_write_queue;
1074 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1075 u32 portid = tsk->portid;
1079 uint mtu, send, sent = 0;
1080 struct iov_iter save;
1082 /* Handle implied connection establishment */
1083 if (unlikely(dest)) {
1084 rc = tipc_sendmsg(iocb, sock, m, dsz);
1085 if (dsz && (dsz == rc))
1086 tsk->sent_unacked = 1;
1089 if (dsz > (uint)INT_MAX)
1095 if (unlikely(sock->state != SS_CONNECTED)) {
1096 if (sock->state == SS_DISCONNECTING)
1103 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1104 dnode = tsk_peer_node(tsk);
1109 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1110 rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1111 if (unlikely(rc < 0))
1114 if (likely(!tsk_conn_cong(tsk))) {
1115 rc = tipc_link_xmit(net, pktchain, dnode, portid);
1117 tsk->sent_unacked++;
1123 if (rc == -EMSGSIZE) {
1124 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1129 if (rc != -ELINKCONG)
1133 rc = tipc_wait_for_sndpkt(sock, &timeo);
1135 __skb_queue_purge(pktchain);
1140 return sent ? sent : rc;
1144 * tipc_send_packet - send a connection-oriented message
1145 * @iocb: if NULL, indicates that socket lock is already held
1146 * @sock: socket structure
1147 * @m: message to send
1148 * @dsz: length of data to be transmitted
1150 * Used for SOCK_SEQPACKET messages.
1152 * Returns the number of bytes sent on success, or errno otherwise
1154 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1155 struct msghdr *m, size_t dsz)
1157 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1160 return tipc_send_stream(iocb, sock, m, dsz);
1163 /* tipc_sk_finish_conn - complete the setup of a connection
1165 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1168 struct sock *sk = &tsk->sk;
1169 struct net *net = sock_net(sk);
1170 struct tipc_msg *msg = &tsk->phdr;
1172 msg_set_destnode(msg, peer_node);
1173 msg_set_destport(msg, peer_port);
1174 msg_set_type(msg, TIPC_CONN_MSG);
1175 msg_set_lookup_scope(msg, 0);
1176 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1178 tsk->probing_intv = CONN_PROBING_INTERVAL;
1179 tsk->probing_state = TIPC_CONN_OK;
1181 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1182 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1183 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1187 * set_orig_addr - capture sender's address for received message
1188 * @m: descriptor for message info
1189 * @msg: received message header
1191 * Note: Address is not captured if not requested by receiver.
1193 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1195 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1198 addr->family = AF_TIPC;
1199 addr->addrtype = TIPC_ADDR_ID;
1200 memset(&addr->addr, 0, sizeof(addr->addr));
1201 addr->addr.id.ref = msg_origport(msg);
1202 addr->addr.id.node = msg_orignode(msg);
1203 addr->addr.name.domain = 0; /* could leave uninitialized */
1204 addr->scope = 0; /* could leave uninitialized */
1205 m->msg_namelen = sizeof(struct sockaddr_tipc);
1210 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1211 * @m: descriptor for message info
1212 * @msg: received message header
1213 * @tsk: TIPC port associated with message
1215 * Note: Ancillary data is not captured if not requested by receiver.
1217 * Returns 0 if successful, otherwise errno
1219 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1220 struct tipc_sock *tsk)
1228 if (likely(m->msg_controllen == 0))
1231 /* Optionally capture errored message object(s) */
1232 err = msg ? msg_errcode(msg) : 0;
1233 if (unlikely(err)) {
1235 anc_data[1] = msg_data_sz(msg);
1236 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1240 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1247 /* Optionally capture message destination object */
1248 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1249 switch (dest_type) {
1250 case TIPC_NAMED_MSG:
1252 anc_data[0] = msg_nametype(msg);
1253 anc_data[1] = msg_namelower(msg);
1254 anc_data[2] = msg_namelower(msg);
1256 case TIPC_MCAST_MSG:
1258 anc_data[0] = msg_nametype(msg);
1259 anc_data[1] = msg_namelower(msg);
1260 anc_data[2] = msg_nameupper(msg);
1263 has_name = (tsk->conn_type != 0);
1264 anc_data[0] = tsk->conn_type;
1265 anc_data[1] = tsk->conn_instance;
1266 anc_data[2] = tsk->conn_instance;
1272 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1280 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1282 struct net *net = sock_net(&tsk->sk);
1283 struct sk_buff *skb = NULL;
1284 struct tipc_msg *msg;
1285 u32 peer_port = tsk_peer_port(tsk);
1286 u32 dnode = tsk_peer_node(tsk);
1288 if (!tsk->connected)
1290 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1291 dnode, tsk_own_node(tsk), peer_port,
1292 tsk->portid, TIPC_OK);
1296 msg_set_msgcnt(msg, ack);
1297 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1300 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1302 struct sock *sk = sock->sk;
1304 long timeo = *timeop;
1308 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1309 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1310 if (sock->state == SS_DISCONNECTING) {
1315 timeo = schedule_timeout(timeo);
1319 if (!skb_queue_empty(&sk->sk_receive_queue))
1321 err = sock_intr_errno(timeo);
1322 if (signal_pending(current))
1328 finish_wait(sk_sleep(sk), &wait);
1334 * tipc_recvmsg - receive packet-oriented message
1336 * @m: descriptor for message info
1337 * @buf_len: total size of user buffer area
1338 * @flags: receive flags
1340 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1341 * If the complete message doesn't fit in user area, truncate it.
1343 * Returns size of returned message data, errno otherwise
1345 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1346 struct msghdr *m, size_t buf_len, int flags)
1348 struct sock *sk = sock->sk;
1349 struct tipc_sock *tsk = tipc_sk(sk);
1350 struct sk_buff *buf;
1351 struct tipc_msg *msg;
1357 /* Catch invalid receive requests */
1358 if (unlikely(!buf_len))
1363 if (unlikely(sock->state == SS_UNCONNECTED)) {
1368 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1371 /* Look for a message in receive queue; wait if necessary */
1372 res = tipc_wait_for_rcvmsg(sock, &timeo);
1376 /* Look at first message in receive queue */
1377 buf = skb_peek(&sk->sk_receive_queue);
1379 sz = msg_data_sz(msg);
1380 err = msg_errcode(msg);
1382 /* Discard an empty non-errored message & try again */
1383 if ((!sz) && (!err)) {
1384 tsk_advance_rx_queue(sk);
1388 /* Capture sender's address (optional) */
1389 set_orig_addr(m, msg);
1391 /* Capture ancillary data (optional) */
1392 res = tipc_sk_anc_data_recv(m, msg, tsk);
1396 /* Capture message data (if valid) & compute return value (always) */
1398 if (unlikely(buf_len < sz)) {
1400 m->msg_flags |= MSG_TRUNC;
1402 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1407 if ((sock->state == SS_READY) ||
1408 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1414 /* Consume received message (optional) */
1415 if (likely(!(flags & MSG_PEEK))) {
1416 if ((sock->state != SS_READY) &&
1417 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1418 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1419 tsk->rcv_unacked = 0;
1421 tsk_advance_rx_queue(sk);
1429 * tipc_recv_stream - receive stream-oriented data
1431 * @m: descriptor for message info
1432 * @buf_len: total size of user buffer area
1433 * @flags: receive flags
1435 * Used for SOCK_STREAM messages only. If not enough data is available
1436 * will optionally wait for more; never truncates data.
1438 * Returns size of returned message data, errno otherwise
1440 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1441 struct msghdr *m, size_t buf_len, int flags)
1443 struct sock *sk = sock->sk;
1444 struct tipc_sock *tsk = tipc_sk(sk);
1445 struct sk_buff *buf;
1446 struct tipc_msg *msg;
1449 int sz_to_copy, target, needed;
1454 /* Catch invalid receive attempts */
1455 if (unlikely(!buf_len))
1460 if (unlikely(sock->state == SS_UNCONNECTED)) {
1465 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1466 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1469 /* Look for a message in receive queue; wait if necessary */
1470 res = tipc_wait_for_rcvmsg(sock, &timeo);
1474 /* Look at first message in receive queue */
1475 buf = skb_peek(&sk->sk_receive_queue);
1477 sz = msg_data_sz(msg);
1478 err = msg_errcode(msg);
1480 /* Discard an empty non-errored message & try again */
1481 if ((!sz) && (!err)) {
1482 tsk_advance_rx_queue(sk);
1486 /* Optionally capture sender's address & ancillary data of first msg */
1487 if (sz_copied == 0) {
1488 set_orig_addr(m, msg);
1489 res = tipc_sk_anc_data_recv(m, msg, tsk);
1494 /* Capture message data (if valid) & compute return value (always) */
1496 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1499 needed = (buf_len - sz_copied);
1500 sz_to_copy = (sz <= needed) ? sz : needed;
1502 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1507 sz_copied += sz_to_copy;
1509 if (sz_to_copy < sz) {
1510 if (!(flags & MSG_PEEK))
1511 TIPC_SKB_CB(buf)->handle =
1512 (void *)(unsigned long)(offset + sz_to_copy);
1517 goto exit; /* can't add error msg to valid data */
1519 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1525 /* Consume received message (optional) */
1526 if (likely(!(flags & MSG_PEEK))) {
1527 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1528 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1529 tsk->rcv_unacked = 0;
1531 tsk_advance_rx_queue(sk);
1534 /* Loop around if more data is required */
1535 if ((sz_copied < buf_len) && /* didn't get all requested data */
1536 (!skb_queue_empty(&sk->sk_receive_queue) ||
1537 (sz_copied < target)) && /* and more is ready or required */
1538 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1539 (!err)) /* and haven't reached a FIN */
1544 return sz_copied ? sz_copied : res;
1548 * tipc_write_space - wake up thread if port congestion is released
1551 static void tipc_write_space(struct sock *sk)
1553 struct socket_wq *wq;
1556 wq = rcu_dereference(sk->sk_wq);
1557 if (wq_has_sleeper(wq))
1558 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1559 POLLWRNORM | POLLWRBAND);
1564 * tipc_data_ready - wake up threads to indicate messages have been received
1566 * @len: the length of messages
1568 static void tipc_data_ready(struct sock *sk)
1570 struct socket_wq *wq;
1573 wq = rcu_dereference(sk->sk_wq);
1574 if (wq_has_sleeper(wq))
1575 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1576 POLLRDNORM | POLLRDBAND);
1581 * filter_connect - Handle all incoming messages for a connection-based socket
1583 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1585 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1587 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1589 struct sock *sk = &tsk->sk;
1590 struct net *net = sock_net(sk);
1591 struct socket *sock = sk->sk_socket;
1592 struct tipc_msg *msg = buf_msg(*skb);
1593 int retval = -TIPC_ERR_NO_PORT;
1598 switch ((int)sock->state) {
1600 /* Accept only connection-based messages sent by peer */
1601 if (tsk_peer_msg(tsk, msg)) {
1602 if (unlikely(msg_errcode(msg))) {
1603 sock->state = SS_DISCONNECTING;
1605 /* let timer expire on it's own */
1606 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1613 /* Accept only ACK or NACK message */
1615 if (unlikely(!msg_connected(msg)))
1618 if (unlikely(msg_errcode(msg))) {
1619 sock->state = SS_DISCONNECTING;
1620 sk->sk_err = ECONNREFUSED;
1625 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1626 sock->state = SS_DISCONNECTING;
1627 sk->sk_err = EINVAL;
1632 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1633 msg_set_importance(&tsk->phdr, msg_importance(msg));
1634 sock->state = SS_CONNECTED;
1636 /* If an incoming message is an 'ACK-', it should be
1637 * discarded here because it doesn't contain useful
1638 * data. In addition, we should try to wake up
1639 * connect() routine if sleeping.
1641 if (msg_data_sz(msg) == 0) {
1644 if (waitqueue_active(sk_sleep(sk)))
1645 wake_up_interruptible(sk_sleep(sk));
1650 case SS_UNCONNECTED:
1651 /* Accept only SYN message */
1652 if (!msg_connected(msg) && !(msg_errcode(msg)))
1655 case SS_DISCONNECTING:
1658 pr_err("Unknown socket state %u\n", sock->state);
1664 * rcvbuf_limit - get proper overload limit of socket receive queue
1668 * For all connection oriented messages, irrespective of importance,
1669 * the default overload value (i.e. 67MB) is set as limit.
1671 * For all connectionless messages, by default new queue limits are
1674 * TIPC_LOW_IMPORTANCE (4 MB)
1675 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1676 * TIPC_HIGH_IMPORTANCE (16 MB)
1677 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1679 * Returns overload limit according to corresponding message importance
1681 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1683 struct tipc_msg *msg = buf_msg(buf);
1685 if (msg_connected(msg))
1686 return sysctl_tipc_rmem[2];
1688 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1689 msg_importance(msg);
1693 * filter_rcv - validate incoming message
1695 * @skb: pointer to message. Set to NULL if buffer is consumed.
1697 * Enqueues message on receive queue if acceptable; optionally handles
1698 * disconnect indication for a connected socket.
1700 * Called with socket lock already taken
1702 * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1704 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1706 struct socket *sock = sk->sk_socket;
1707 struct tipc_sock *tsk = tipc_sk(sk);
1708 struct tipc_msg *msg = buf_msg(*skb);
1709 unsigned int limit = rcvbuf_limit(sk, *skb);
1712 if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1713 tipc_sk_proto_rcv(tsk, skb);
1717 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1720 sk->sk_write_space(sk);
1725 /* Reject message if it is wrong sort of message for socket */
1726 if (msg_type(msg) > TIPC_DIRECT_MSG)
1727 return -TIPC_ERR_NO_PORT;
1729 if (sock->state == SS_READY) {
1730 if (msg_connected(msg))
1731 return -TIPC_ERR_NO_PORT;
1733 rc = filter_connect(tsk, skb);
1734 if (rc != TIPC_OK || !*skb)
1738 /* Reject message if there isn't room to queue it */
1739 if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1740 return -TIPC_ERR_OVERLOAD;
1742 /* Enqueue message */
1743 TIPC_SKB_CB(*skb)->handle = NULL;
1744 __skb_queue_tail(&sk->sk_receive_queue, *skb);
1745 skb_set_owner_r(*skb, sk);
1747 sk->sk_data_ready(sk);
1753 * tipc_backlog_rcv - handle incoming message from backlog queue
1757 * Caller must hold socket lock
1761 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1766 struct tipc_sock *tsk = tipc_sk(sk);
1767 struct net *net = sock_net(sk);
1768 uint truesize = skb->truesize;
1770 err = filter_rcv(sk, &skb);
1772 dcnt = &tsk->dupl_rcvcnt;
1773 if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1774 atomic_add(truesize, dcnt);
1777 if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1778 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
1783 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1784 * inputq and try adding them to socket or backlog queue
1785 * @inputq: list of incoming buffers with potentially different destinations
1786 * @sk: socket where the buffers should be enqueued
1787 * @dport: port number for the socket
1788 * @_skb: returned buffer to be forwarded or rejected, if applicable
1790 * Caller must hold socket lock
1792 * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1793 * or -TIPC_ERR_NO_PORT
1795 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1796 u32 dport, struct sk_buff **_skb)
1801 struct sk_buff *skb;
1802 unsigned long time_limit = jiffies + 2;
1804 while (skb_queue_len(inputq)) {
1805 skb = tipc_skb_dequeue(inputq, dport);
1808 /* Return if softirq window exhausted */
1809 if (unlikely(time_after_eq(jiffies, time_limit)))
1811 if (!sock_owned_by_user(sk)) {
1812 err = filter_rcv(sk, &skb);
1818 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1819 if (sk->sk_backlog.len)
1820 atomic_set(dcnt, 0);
1821 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1822 if (likely(!sk_add_backlog(sk, skb, lim)))
1825 return -TIPC_ERR_OVERLOAD;
1831 * tipc_sk_rcv - handle a chain of incoming buffers
1832 * @inputq: buffer list containing the buffers
1833 * Consumes all buffers in list until inputq is empty
1834 * Note: may be called in multiple threads referring to the same queue
1835 * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1836 * Only node local calls check the return value, sending single-buffer queues
1838 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1840 u32 dnode, dport = 0;
1841 int err = -TIPC_ERR_NO_PORT;
1842 struct sk_buff *skb;
1843 struct tipc_sock *tsk;
1844 struct tipc_net *tn;
1847 while (skb_queue_len(inputq)) {
1849 dport = tipc_skb_peek_port(inputq, dport);
1850 tsk = tipc_sk_lookup(net, dport);
1853 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1854 err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1855 spin_unlock_bh(&sk->sk_lock.slock);
1860 skb = tipc_skb_dequeue(inputq, dport);
1864 if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1867 dnode = msg_destnode(buf_msg(skb));
1870 tn = net_generic(net, tipc_net_id);
1871 if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1874 tipc_link_xmit_skb(net, skb, dnode, dport);
1876 return err ? -EHOSTUNREACH : 0;
1879 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1881 struct sock *sk = sock->sk;
1886 int err = sock_error(sk);
1891 if (signal_pending(current))
1892 return sock_intr_errno(*timeo_p);
1894 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1895 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1896 finish_wait(sk_sleep(sk), &wait);
1902 * tipc_connect - establish a connection to another TIPC port
1903 * @sock: socket structure
1904 * @dest: socket address for destination port
1905 * @destlen: size of socket address data structure
1906 * @flags: file-related flags associated with socket
1908 * Returns 0 on success, errno otherwise
1910 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1911 int destlen, int flags)
1913 struct sock *sk = sock->sk;
1914 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1915 struct msghdr m = {NULL,};
1916 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1917 socket_state previous;
1922 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1923 if (sock->state == SS_READY) {
1929 * Reject connection attempt using multicast address
1931 * Note: send_msg() validates the rest of the address fields,
1932 * so there's no need to do it here
1934 if (dst->addrtype == TIPC_ADDR_MCAST) {
1939 previous = sock->state;
1940 switch (sock->state) {
1941 case SS_UNCONNECTED:
1942 /* Send a 'SYN-' to destination */
1944 m.msg_namelen = destlen;
1946 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1947 * indicate send_msg() is never blocked.
1950 m.msg_flags = MSG_DONTWAIT;
1952 res = tipc_sendmsg(NULL, sock, &m, 0);
1953 if ((res < 0) && (res != -EWOULDBLOCK))
1956 /* Just entered SS_CONNECTING state; the only
1957 * difference is that return value in non-blocking
1958 * case is EINPROGRESS, rather than EALREADY.
1962 if (previous == SS_CONNECTING)
1966 timeout = msecs_to_jiffies(timeout);
1967 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1968 res = tipc_wait_for_connect(sock, &timeout);
1983 * tipc_listen - allow socket to listen for incoming connections
1984 * @sock: socket structure
1987 * Returns 0 on success, errno otherwise
1989 static int tipc_listen(struct socket *sock, int len)
1991 struct sock *sk = sock->sk;
1996 if (sock->state != SS_UNCONNECTED)
1999 sock->state = SS_LISTENING;
2007 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2009 struct sock *sk = sock->sk;
2013 /* True wake-one mechanism for incoming connections: only
2014 * one process gets woken up, not the 'whole herd'.
2015 * Since we do not 'race & poll' for established sockets
2016 * anymore, the common case will execute the loop only once.
2019 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2020 TASK_INTERRUPTIBLE);
2021 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2023 timeo = schedule_timeout(timeo);
2027 if (!skb_queue_empty(&sk->sk_receive_queue))
2030 if (sock->state != SS_LISTENING)
2032 err = sock_intr_errno(timeo);
2033 if (signal_pending(current))
2039 finish_wait(sk_sleep(sk), &wait);
2044 * tipc_accept - wait for connection request
2045 * @sock: listening socket
2046 * @newsock: new socket that is to be connected
2047 * @flags: file-related flags associated with socket
2049 * Returns 0 on success, errno otherwise
2051 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2053 struct sock *new_sk, *sk = sock->sk;
2054 struct sk_buff *buf;
2055 struct tipc_sock *new_tsock;
2056 struct tipc_msg *msg;
2062 if (sock->state != SS_LISTENING) {
2066 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2067 res = tipc_wait_for_accept(sock, timeo);
2071 buf = skb_peek(&sk->sk_receive_queue);
2073 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2077 new_sk = new_sock->sk;
2078 new_tsock = tipc_sk(new_sk);
2081 /* we lock on new_sk; but lockdep sees the lock on sk */
2082 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2085 * Reject any stray messages received by new socket
2086 * before the socket lock was taken (very, very unlikely)
2088 tsk_rej_rx_queue(new_sk);
2090 /* Connect new socket to it's peer */
2091 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2092 new_sock->state = SS_CONNECTED;
2094 tsk_set_importance(new_tsock, msg_importance(msg));
2095 if (msg_named(msg)) {
2096 new_tsock->conn_type = msg_nametype(msg);
2097 new_tsock->conn_instance = msg_nameinst(msg);
2101 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2102 * Respond to 'SYN+' by queuing it on new socket.
2104 if (!msg_data_sz(msg)) {
2105 struct msghdr m = {NULL,};
2107 tsk_advance_rx_queue(sk);
2108 tipc_send_packet(NULL, new_sock, &m, 0);
2110 __skb_dequeue(&sk->sk_receive_queue);
2111 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2112 skb_set_owner_r(buf, new_sk);
2114 release_sock(new_sk);
2121 * tipc_shutdown - shutdown socket connection
2122 * @sock: socket structure
2123 * @how: direction to close (must be SHUT_RDWR)
2125 * Terminates connection (if necessary), then purges socket's receive queue.
2127 * Returns 0 on success, errno otherwise
2129 static int tipc_shutdown(struct socket *sock, int how)
2131 struct sock *sk = sock->sk;
2132 struct net *net = sock_net(sk);
2133 struct tipc_sock *tsk = tipc_sk(sk);
2134 struct sk_buff *skb;
2138 if (how != SHUT_RDWR)
2143 switch (sock->state) {
2148 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2149 skb = __skb_dequeue(&sk->sk_receive_queue);
2151 if (TIPC_SKB_CB(skb)->handle != NULL) {
2155 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2156 TIPC_CONN_SHUTDOWN))
2157 tipc_link_xmit_skb(net, skb, dnode,
2159 tipc_node_remove_conn(net, dnode, tsk->portid);
2161 dnode = tsk_peer_node(tsk);
2163 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2164 TIPC_CONN_MSG, SHORT_H_SIZE,
2165 0, dnode, tsk_own_node(tsk),
2167 tsk->portid, TIPC_CONN_SHUTDOWN);
2168 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2171 sock->state = SS_DISCONNECTING;
2172 tipc_node_remove_conn(net, dnode, tsk->portid);
2175 case SS_DISCONNECTING:
2177 /* Discard any unreceived messages */
2178 __skb_queue_purge(&sk->sk_receive_queue);
2180 /* Wake up anyone sleeping in poll */
2181 sk->sk_state_change(sk);
2193 static void tipc_sk_timeout(unsigned long data)
2195 struct tipc_sock *tsk = (struct tipc_sock *)data;
2196 struct sock *sk = &tsk->sk;
2197 struct sk_buff *skb = NULL;
2198 u32 peer_port, peer_node;
2199 u32 own_node = tsk_own_node(tsk);
2202 if (!tsk->connected) {
2206 peer_port = tsk_peer_port(tsk);
2207 peer_node = tsk_peer_node(tsk);
2209 if (tsk->probing_state == TIPC_CONN_PROBING) {
2210 /* Previous probe not answered -> self abort */
2211 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2212 TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2213 own_node, peer_node, tsk->portid,
2214 peer_port, TIPC_ERR_NO_PORT);
2216 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2217 INT_H_SIZE, 0, peer_node, own_node,
2218 peer_port, tsk->portid, TIPC_OK);
2219 tsk->probing_state = TIPC_CONN_PROBING;
2220 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2224 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2229 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2230 struct tipc_name_seq const *seq)
2232 struct net *net = sock_net(&tsk->sk);
2233 struct publication *publ;
2238 key = tsk->portid + tsk->pub_count + 1;
2239 if (key == tsk->portid)
2242 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2243 scope, tsk->portid, key);
2244 if (unlikely(!publ))
2247 list_add(&publ->pport_list, &tsk->publications);
2253 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2254 struct tipc_name_seq const *seq)
2256 struct net *net = sock_net(&tsk->sk);
2257 struct publication *publ;
2258 struct publication *safe;
2261 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2263 if (publ->scope != scope)
2265 if (publ->type != seq->type)
2267 if (publ->lower != seq->lower)
2269 if (publ->upper != seq->upper)
2271 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2272 publ->ref, publ->key);
2276 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2277 publ->ref, publ->key);
2280 if (list_empty(&tsk->publications))
2285 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2286 int len, int full_id)
2288 struct net *net = sock_net(&tsk->sk);
2289 struct tipc_net *tn = net_generic(net, tipc_net_id);
2290 struct publication *publ;
2294 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2295 tipc_zone(tn->own_addr),
2296 tipc_cluster(tn->own_addr),
2297 tipc_node(tn->own_addr), tsk->portid);
2299 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2301 if (tsk->connected) {
2302 u32 dport = tsk_peer_port(tsk);
2303 u32 destnode = tsk_peer_node(tsk);
2305 ret += tipc_snprintf(buf + ret, len - ret,
2306 " connected to <%u.%u.%u:%u>",
2307 tipc_zone(destnode),
2308 tipc_cluster(destnode),
2309 tipc_node(destnode), dport);
2310 if (tsk->conn_type != 0)
2311 ret += tipc_snprintf(buf + ret, len - ret,
2312 " via {%u,%u}", tsk->conn_type,
2313 tsk->conn_instance);
2314 } else if (tsk->published) {
2315 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2316 list_for_each_entry(publ, &tsk->publications, pport_list) {
2317 if (publ->lower == publ->upper)
2318 ret += tipc_snprintf(buf + ret, len - ret,
2319 " {%u,%u}", publ->type,
2322 ret += tipc_snprintf(buf + ret, len - ret,
2323 " {%u,%u,%u}", publ->type,
2324 publ->lower, publ->upper);
2327 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2331 struct sk_buff *tipc_sk_socks_show(struct net *net)
2333 struct tipc_net *tn = net_generic(net, tipc_net_id);
2334 const struct bucket_table *tbl;
2335 struct rhash_head *pos;
2336 struct sk_buff *buf;
2337 struct tlv_desc *rep_tlv;
2340 struct tipc_sock *tsk;
2344 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2347 rep_tlv = (struct tlv_desc *)buf->data;
2348 pb = TLV_DATA(rep_tlv);
2349 pb_len = ULTRA_STRING_MAX_LEN;
2352 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2353 for (i = 0; i < tbl->size; i++) {
2354 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2355 spin_lock_bh(&tsk->sk.sk_lock.slock);
2356 str_len += tipc_sk_show(tsk, pb + str_len,
2357 pb_len - str_len, 0);
2358 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2363 str_len += 1; /* for "\0" */
2364 skb_put(buf, TLV_SPACE(str_len));
2365 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2370 /* tipc_sk_reinit: set non-zero address in all existing sockets
2371 * when we go from standalone to network mode.
2373 void tipc_sk_reinit(struct net *net)
2375 struct tipc_net *tn = net_generic(net, tipc_net_id);
2376 const struct bucket_table *tbl;
2377 struct rhash_head *pos;
2378 struct tipc_sock *tsk;
2379 struct tipc_msg *msg;
2383 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2384 for (i = 0; i < tbl->size; i++) {
2385 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2386 spin_lock_bh(&tsk->sk.sk_lock.slock);
2388 msg_set_prevnode(msg, tn->own_addr);
2389 msg_set_orignode(msg, tn->own_addr);
2390 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2396 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2398 struct tipc_net *tn = net_generic(net, tipc_net_id);
2399 struct tipc_sock *tsk;
2402 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2404 sock_hold(&tsk->sk);
2410 static int tipc_sk_insert(struct tipc_sock *tsk)
2412 struct sock *sk = &tsk->sk;
2413 struct net *net = sock_net(sk);
2414 struct tipc_net *tn = net_generic(net, tipc_net_id);
2415 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2416 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2418 while (remaining--) {
2420 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2421 portid = TIPC_MIN_PORT;
2422 tsk->portid = portid;
2423 sock_hold(&tsk->sk);
2424 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2432 static void tipc_sk_remove(struct tipc_sock *tsk)
2434 struct sock *sk = &tsk->sk;
2435 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2437 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2438 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2443 int tipc_sk_rht_init(struct net *net)
2445 struct tipc_net *tn = net_generic(net, tipc_net_id);
2446 struct rhashtable_params rht_params = {
2448 .head_offset = offsetof(struct tipc_sock, node),
2449 .key_offset = offsetof(struct tipc_sock, portid),
2450 .key_len = sizeof(u32), /* portid */
2452 .max_shift = 20, /* 1M */
2453 .min_shift = 8, /* 256 */
2454 .grow_decision = rht_grow_above_75,
2455 .shrink_decision = rht_shrink_below_30,
2458 return rhashtable_init(&tn->sk_rht, &rht_params);
2461 void tipc_sk_rht_destroy(struct net *net)
2463 struct tipc_net *tn = net_generic(net, tipc_net_id);
2465 /* Wait for socket readers to complete */
2468 rhashtable_destroy(&tn->sk_rht);
2472 * tipc_setsockopt - set socket option
2473 * @sock: socket structure
2474 * @lvl: option level
2475 * @opt: option identifier
2476 * @ov: pointer to new option value
2477 * @ol: length of option value
2479 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2480 * (to ease compatibility).
2482 * Returns 0 on success, errno otherwise
2484 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2485 char __user *ov, unsigned int ol)
2487 struct sock *sk = sock->sk;
2488 struct tipc_sock *tsk = tipc_sk(sk);
2492 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2494 if (lvl != SOL_TIPC)
2495 return -ENOPROTOOPT;
2496 if (ol < sizeof(value))
2498 res = get_user(value, (u32 __user *)ov);
2505 case TIPC_IMPORTANCE:
2506 res = tsk_set_importance(tsk, value);
2508 case TIPC_SRC_DROPPABLE:
2509 if (sock->type != SOCK_STREAM)
2510 tsk_set_unreliable(tsk, value);
2514 case TIPC_DEST_DROPPABLE:
2515 tsk_set_unreturnable(tsk, value);
2517 case TIPC_CONN_TIMEOUT:
2518 tipc_sk(sk)->conn_timeout = value;
2519 /* no need to set "res", since already 0 at this point */
2531 * tipc_getsockopt - get socket option
2532 * @sock: socket structure
2533 * @lvl: option level
2534 * @opt: option identifier
2535 * @ov: receptacle for option value
2536 * @ol: receptacle for length of option value
2538 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2539 * (to ease compatibility).
2541 * Returns 0 on success, errno otherwise
2543 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2544 char __user *ov, int __user *ol)
2546 struct sock *sk = sock->sk;
2547 struct tipc_sock *tsk = tipc_sk(sk);
2552 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2553 return put_user(0, ol);
2554 if (lvl != SOL_TIPC)
2555 return -ENOPROTOOPT;
2556 res = get_user(len, ol);
2563 case TIPC_IMPORTANCE:
2564 value = tsk_importance(tsk);
2566 case TIPC_SRC_DROPPABLE:
2567 value = tsk_unreliable(tsk);
2569 case TIPC_DEST_DROPPABLE:
2570 value = tsk_unreturnable(tsk);
2572 case TIPC_CONN_TIMEOUT:
2573 value = tsk->conn_timeout;
2574 /* no need to set "res", since already 0 at this point */
2576 case TIPC_NODE_RECVQ_DEPTH:
2577 value = 0; /* was tipc_queue_size, now obsolete */
2579 case TIPC_SOCK_RECVQ_DEPTH:
2580 value = skb_queue_len(&sk->sk_receive_queue);
2589 return res; /* "get" failed */
2591 if (len < sizeof(value))
2594 if (copy_to_user(ov, &value, sizeof(value)))
2597 return put_user(sizeof(value), ol);
2600 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2602 struct sock *sk = sock->sk;
2603 struct tipc_sioc_ln_req lnr;
2604 void __user *argp = (void __user *)arg;
2607 case SIOCGETLINKNAME:
2608 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2610 if (!tipc_node_get_linkname(sock_net(sk),
2611 lnr.bearer_id & 0xffff, lnr.peer,
2612 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2613 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2617 return -EADDRNOTAVAIL;
2619 return -ENOIOCTLCMD;
2623 /* Protocol switches for the various types of TIPC sockets */
2625 static const struct proto_ops msg_ops = {
2626 .owner = THIS_MODULE,
2628 .release = tipc_release,
2630 .connect = tipc_connect,
2631 .socketpair = sock_no_socketpair,
2632 .accept = sock_no_accept,
2633 .getname = tipc_getname,
2635 .ioctl = tipc_ioctl,
2636 .listen = sock_no_listen,
2637 .shutdown = tipc_shutdown,
2638 .setsockopt = tipc_setsockopt,
2639 .getsockopt = tipc_getsockopt,
2640 .sendmsg = tipc_sendmsg,
2641 .recvmsg = tipc_recvmsg,
2642 .mmap = sock_no_mmap,
2643 .sendpage = sock_no_sendpage
2646 static const struct proto_ops packet_ops = {
2647 .owner = THIS_MODULE,
2649 .release = tipc_release,
2651 .connect = tipc_connect,
2652 .socketpair = sock_no_socketpair,
2653 .accept = tipc_accept,
2654 .getname = tipc_getname,
2656 .ioctl = tipc_ioctl,
2657 .listen = tipc_listen,
2658 .shutdown = tipc_shutdown,
2659 .setsockopt = tipc_setsockopt,
2660 .getsockopt = tipc_getsockopt,
2661 .sendmsg = tipc_send_packet,
2662 .recvmsg = tipc_recvmsg,
2663 .mmap = sock_no_mmap,
2664 .sendpage = sock_no_sendpage
2667 static const struct proto_ops stream_ops = {
2668 .owner = THIS_MODULE,
2670 .release = tipc_release,
2672 .connect = tipc_connect,
2673 .socketpair = sock_no_socketpair,
2674 .accept = tipc_accept,
2675 .getname = tipc_getname,
2677 .ioctl = tipc_ioctl,
2678 .listen = tipc_listen,
2679 .shutdown = tipc_shutdown,
2680 .setsockopt = tipc_setsockopt,
2681 .getsockopt = tipc_getsockopt,
2682 .sendmsg = tipc_send_stream,
2683 .recvmsg = tipc_recv_stream,
2684 .mmap = sock_no_mmap,
2685 .sendpage = sock_no_sendpage
2688 static const struct net_proto_family tipc_family_ops = {
2689 .owner = THIS_MODULE,
2691 .create = tipc_sk_create
2694 static struct proto tipc_proto = {
2696 .owner = THIS_MODULE,
2697 .obj_size = sizeof(struct tipc_sock),
2698 .sysctl_rmem = sysctl_tipc_rmem
2701 static struct proto tipc_proto_kern = {
2703 .obj_size = sizeof(struct tipc_sock),
2704 .sysctl_rmem = sysctl_tipc_rmem
2708 * tipc_socket_init - initialize TIPC socket interface
2710 * Returns 0 on success, errno otherwise
2712 int tipc_socket_init(void)
2716 res = proto_register(&tipc_proto, 1);
2718 pr_err("Failed to register TIPC protocol type\n");
2722 res = sock_register(&tipc_family_ops);
2724 pr_err("Failed to register TIPC socket type\n");
2725 proto_unregister(&tipc_proto);
2733 * tipc_socket_stop - stop TIPC socket interface
2735 void tipc_socket_stop(void)
2737 sock_unregister(tipc_family_ops.family);
2738 proto_unregister(&tipc_proto);
2741 /* Caller should hold socket lock for the passed tipc socket. */
2742 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2746 struct nlattr *nest;
2748 peer_node = tsk_peer_node(tsk);
2749 peer_port = tsk_peer_port(tsk);
2751 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2753 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2755 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2758 if (tsk->conn_type != 0) {
2759 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2761 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2763 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2766 nla_nest_end(skb, nest);
2771 nla_nest_cancel(skb, nest);
2776 /* Caller should hold socket lock for the passed tipc socket. */
2777 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2778 struct tipc_sock *tsk)
2782 struct nlattr *attrs;
2783 struct net *net = sock_net(skb->sk);
2784 struct tipc_net *tn = net_generic(net, tipc_net_id);
2786 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2787 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2791 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2793 goto genlmsg_cancel;
2794 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2795 goto attr_msg_cancel;
2796 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2797 goto attr_msg_cancel;
2799 if (tsk->connected) {
2800 err = __tipc_nl_add_sk_con(skb, tsk);
2802 goto attr_msg_cancel;
2803 } else if (!list_empty(&tsk->publications)) {
2804 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2805 goto attr_msg_cancel;
2807 nla_nest_end(skb, attrs);
2808 genlmsg_end(skb, hdr);
2813 nla_nest_cancel(skb, attrs);
2815 genlmsg_cancel(skb, hdr);
2820 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2823 struct tipc_sock *tsk;
2824 const struct bucket_table *tbl;
2825 struct rhash_head *pos;
2826 struct net *net = sock_net(skb->sk);
2827 struct tipc_net *tn = net_generic(net, tipc_net_id);
2828 u32 tbl_id = cb->args[0];
2829 u32 prev_portid = cb->args[1];
2832 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2833 for (; tbl_id < tbl->size; tbl_id++) {
2834 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2835 spin_lock_bh(&tsk->sk.sk_lock.slock);
2836 if (prev_portid && prev_portid != tsk->portid) {
2837 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2841 err = __tipc_nl_add_sk(skb, cb, tsk);
2843 prev_portid = tsk->portid;
2844 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2848 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2853 cb->args[0] = tbl_id;
2854 cb->args[1] = prev_portid;
2859 /* Caller should hold socket lock for the passed tipc socket. */
2860 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2861 struct netlink_callback *cb,
2862 struct publication *publ)
2865 struct nlattr *attrs;
2867 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2868 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2872 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2874 goto genlmsg_cancel;
2876 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2877 goto attr_msg_cancel;
2878 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2879 goto attr_msg_cancel;
2880 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2881 goto attr_msg_cancel;
2882 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2883 goto attr_msg_cancel;
2885 nla_nest_end(skb, attrs);
2886 genlmsg_end(skb, hdr);
2891 nla_nest_cancel(skb, attrs);
2893 genlmsg_cancel(skb, hdr);
2898 /* Caller should hold socket lock for the passed tipc socket. */
2899 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2900 struct netlink_callback *cb,
2901 struct tipc_sock *tsk, u32 *last_publ)
2904 struct publication *p;
2907 list_for_each_entry(p, &tsk->publications, pport_list) {
2908 if (p->key == *last_publ)
2911 if (p->key != *last_publ) {
2912 /* We never set seq or call nl_dump_check_consistent()
2913 * this means that setting prev_seq here will cause the
2914 * consistence check to fail in the netlink callback
2915 * handler. Resulting in the last NLMSG_DONE message
2916 * having the NLM_F_DUMP_INTR flag set.
2923 p = list_first_entry(&tsk->publications, struct publication,
2927 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2928 err = __tipc_nl_add_sk_publ(skb, cb, p);
2930 *last_publ = p->key;
2939 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2942 u32 tsk_portid = cb->args[0];
2943 u32 last_publ = cb->args[1];
2944 u32 done = cb->args[2];
2945 struct net *net = sock_net(skb->sk);
2946 struct tipc_sock *tsk;
2949 struct nlattr **attrs;
2950 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2952 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2956 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2957 attrs[TIPC_NLA_SOCK],
2958 tipc_nl_sock_policy);
2962 if (!sock[TIPC_NLA_SOCK_REF])
2965 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2971 tsk = tipc_sk_lookup(net, tsk_portid);
2975 lock_sock(&tsk->sk);
2976 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2979 release_sock(&tsk->sk);
2982 cb->args[0] = tsk_portid;
2983 cb->args[1] = last_publ;