2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2014, 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);
779 /* tipc_sk_mcast_rcv - Deliver multicast message to all destination sockets
781 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff *buf)
783 struct tipc_msg *msg = buf_msg(buf);
784 struct tipc_port_list dports = {0, NULL, };
785 struct tipc_port_list *item;
787 uint i, last, dst = 0;
788 u32 scope = TIPC_CLUSTER_SCOPE;
789 struct sk_buff_head msgs;
791 if (in_own_node(net, msg_orignode(msg)))
792 scope = TIPC_NODE_SCOPE;
794 if (unlikely(!msg_mcast(msg))) {
795 pr_warn("Received non-multicast msg in multicast\n");
799 /* Create destination port list: */
800 tipc_nametbl_mc_translate(net, msg_nametype(msg), msg_namelower(msg),
801 msg_nameupper(msg), scope, &dports);
808 for (item = &dports; item; item = item->next) {
809 for (i = 0; i < PLSIZE && ++dst <= last; i++) {
810 b = (dst != last) ? skb_clone(buf, GFP_ATOMIC) : buf;
812 pr_warn("Failed do clone mcast rcv buffer\n");
815 msg_set_destport(msg, item->ports[i]);
816 skb_queue_head_init(&msgs);
817 skb_queue_tail(&msgs, b);
818 tipc_sk_rcv(net, &msgs);
822 tipc_port_list_free(&dports);
826 * tipc_sk_proto_rcv - receive a connection mng protocol message
827 * @tsk: receiving socket
828 * @skb: pointer to message buffer. Set to NULL if buffer is consumed.
830 static void tipc_sk_proto_rcv(struct tipc_sock *tsk, struct sk_buff **skb)
832 struct tipc_msg *msg = buf_msg(*skb);
835 u32 own_node = tsk_own_node(tsk);
836 /* Ignore if connection cannot be validated: */
837 if (!tsk_peer_msg(tsk, msg))
840 tsk->probing_state = TIPC_CONN_OK;
842 if (msg_type(msg) == CONN_ACK) {
843 conn_cong = tsk_conn_cong(tsk);
844 tsk->sent_unacked -= msg_msgcnt(msg);
846 tsk->sk.sk_write_space(&tsk->sk);
847 } else if (msg_type(msg) == CONN_PROBE) {
848 if (tipc_msg_reverse(own_node, *skb, &dnode, TIPC_OK)) {
849 msg_set_type(msg, CONN_PROBE_REPLY);
853 /* Do nothing if msg_type() == CONN_PROBE_REPLY */
859 static int tipc_wait_for_sndmsg(struct socket *sock, long *timeo_p)
861 struct sock *sk = sock->sk;
862 struct tipc_sock *tsk = tipc_sk(sk);
867 int err = sock_error(sk);
870 if (sock->state == SS_DISCONNECTING)
874 if (signal_pending(current))
875 return sock_intr_errno(*timeo_p);
877 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
878 done = sk_wait_event(sk, timeo_p, !tsk->link_cong);
879 finish_wait(sk_sleep(sk), &wait);
885 * tipc_sendmsg - send message in connectionless manner
886 * @iocb: if NULL, indicates that socket lock is already held
887 * @sock: socket structure
888 * @m: message to send
889 * @dsz: amount of user data to be sent
891 * Message must have an destination specified explicitly.
892 * Used for SOCK_RDM and SOCK_DGRAM messages,
893 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
894 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
896 * Returns the number of bytes sent on success, or errno otherwise
898 static int tipc_sendmsg(struct kiocb *iocb, struct socket *sock,
899 struct msghdr *m, size_t dsz)
901 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
902 struct sock *sk = sock->sk;
903 struct tipc_sock *tsk = tipc_sk(sk);
904 struct net *net = sock_net(sk);
905 struct tipc_msg *mhdr = &tsk->phdr;
907 struct sk_buff_head *pktchain = &sk->sk_write_queue;
909 struct tipc_name_seq *seq = &dest->addr.nameseq;
910 struct iov_iter save;
916 return -EDESTADDRREQ;
918 if (unlikely((m->msg_namelen < sizeof(*dest)) ||
919 (dest->family != AF_TIPC)))
922 if (dsz > TIPC_MAX_USER_MSG_SIZE)
928 if (unlikely(sock->state != SS_READY)) {
929 if (sock->state == SS_LISTENING) {
933 if (sock->state != SS_UNCONNECTED) {
937 if (tsk->published) {
941 if (dest->addrtype == TIPC_ADDR_NAME) {
942 tsk->conn_type = dest->addr.name.name.type;
943 tsk->conn_instance = dest->addr.name.name.instance;
947 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
949 if (dest->addrtype == TIPC_ADDR_MCAST) {
950 rc = tipc_sendmcast(sock, seq, m, dsz, timeo);
952 } else if (dest->addrtype == TIPC_ADDR_NAME) {
953 u32 type = dest->addr.name.name.type;
954 u32 inst = dest->addr.name.name.instance;
955 u32 domain = dest->addr.name.domain;
958 msg_set_type(mhdr, TIPC_NAMED_MSG);
959 msg_set_hdr_sz(mhdr, NAMED_H_SIZE);
960 msg_set_nametype(mhdr, type);
961 msg_set_nameinst(mhdr, inst);
962 msg_set_lookup_scope(mhdr, tipc_addr_scope(domain));
963 dport = tipc_nametbl_translate(net, type, inst, &dnode);
964 msg_set_destnode(mhdr, dnode);
965 msg_set_destport(mhdr, dport);
966 if (unlikely(!dport && !dnode)) {
970 } else if (dest->addrtype == TIPC_ADDR_ID) {
971 dnode = dest->addr.id.node;
972 msg_set_type(mhdr, TIPC_DIRECT_MSG);
973 msg_set_lookup_scope(mhdr, 0);
974 msg_set_destnode(mhdr, dnode);
975 msg_set_destport(mhdr, dest->addr.id.ref);
976 msg_set_hdr_sz(mhdr, BASIC_H_SIZE);
981 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
982 rc = tipc_msg_build(mhdr, m, 0, dsz, mtu, pktchain);
987 skb = skb_peek(pktchain);
988 TIPC_SKB_CB(skb)->wakeup_pending = tsk->link_cong;
989 rc = tipc_link_xmit(net, pktchain, dnode, tsk->portid);
990 if (likely(rc >= 0)) {
991 if (sock->state != SS_READY)
992 sock->state = SS_CONNECTING;
996 if (rc == -EMSGSIZE) {
1000 if (rc != -ELINKCONG)
1003 rc = tipc_wait_for_sndmsg(sock, &timeo);
1005 __skb_queue_purge(pktchain);
1014 static int tipc_wait_for_sndpkt(struct socket *sock, long *timeo_p)
1016 struct sock *sk = sock->sk;
1017 struct tipc_sock *tsk = tipc_sk(sk);
1022 int err = sock_error(sk);
1025 if (sock->state == SS_DISCONNECTING)
1027 else if (sock->state != SS_CONNECTED)
1031 if (signal_pending(current))
1032 return sock_intr_errno(*timeo_p);
1034 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1035 done = sk_wait_event(sk, timeo_p,
1037 !tsk_conn_cong(tsk)) ||
1039 finish_wait(sk_sleep(sk), &wait);
1045 * tipc_send_stream - send stream-oriented data
1047 * @sock: socket structure
1049 * @dsz: total length of data to be transmitted
1051 * Used for SOCK_STREAM data.
1053 * Returns the number of bytes sent on success (or partial success),
1054 * or errno if no data sent
1056 static int tipc_send_stream(struct kiocb *iocb, struct socket *sock,
1057 struct msghdr *m, size_t dsz)
1059 struct sock *sk = sock->sk;
1060 struct net *net = sock_net(sk);
1061 struct tipc_sock *tsk = tipc_sk(sk);
1062 struct tipc_msg *mhdr = &tsk->phdr;
1063 struct sk_buff_head *pktchain = &sk->sk_write_queue;
1064 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1065 u32 portid = tsk->portid;
1069 uint mtu, send, sent = 0;
1070 struct iov_iter save;
1072 /* Handle implied connection establishment */
1073 if (unlikely(dest)) {
1074 rc = tipc_sendmsg(iocb, sock, m, dsz);
1075 if (dsz && (dsz == rc))
1076 tsk->sent_unacked = 1;
1079 if (dsz > (uint)INT_MAX)
1085 if (unlikely(sock->state != SS_CONNECTED)) {
1086 if (sock->state == SS_DISCONNECTING)
1093 timeo = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1094 dnode = tsk_peer_node(tsk);
1099 send = min_t(uint, dsz - sent, TIPC_MAX_USER_MSG_SIZE);
1100 rc = tipc_msg_build(mhdr, m, sent, send, mtu, pktchain);
1101 if (unlikely(rc < 0))
1104 if (likely(!tsk_conn_cong(tsk))) {
1105 rc = tipc_link_xmit(net, pktchain, dnode, portid);
1107 tsk->sent_unacked++;
1113 if (rc == -EMSGSIZE) {
1114 tsk->max_pkt = tipc_node_get_mtu(net, dnode,
1119 if (rc != -ELINKCONG)
1123 rc = tipc_wait_for_sndpkt(sock, &timeo);
1125 __skb_queue_purge(pktchain);
1130 return sent ? sent : rc;
1134 * tipc_send_packet - send a connection-oriented message
1135 * @iocb: if NULL, indicates that socket lock is already held
1136 * @sock: socket structure
1137 * @m: message to send
1138 * @dsz: length of data to be transmitted
1140 * Used for SOCK_SEQPACKET messages.
1142 * Returns the number of bytes sent on success, or errno otherwise
1144 static int tipc_send_packet(struct kiocb *iocb, struct socket *sock,
1145 struct msghdr *m, size_t dsz)
1147 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1150 return tipc_send_stream(iocb, sock, m, dsz);
1153 /* tipc_sk_finish_conn - complete the setup of a connection
1155 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1158 struct sock *sk = &tsk->sk;
1159 struct net *net = sock_net(sk);
1160 struct tipc_msg *msg = &tsk->phdr;
1162 msg_set_destnode(msg, peer_node);
1163 msg_set_destport(msg, peer_port);
1164 msg_set_type(msg, TIPC_CONN_MSG);
1165 msg_set_lookup_scope(msg, 0);
1166 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1168 tsk->probing_intv = CONN_PROBING_INTERVAL;
1169 tsk->probing_state = TIPC_CONN_OK;
1171 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
1172 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1173 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1177 * set_orig_addr - capture sender's address for received message
1178 * @m: descriptor for message info
1179 * @msg: received message header
1181 * Note: Address is not captured if not requested by receiver.
1183 static void set_orig_addr(struct msghdr *m, struct tipc_msg *msg)
1185 DECLARE_SOCKADDR(struct sockaddr_tipc *, addr, m->msg_name);
1188 addr->family = AF_TIPC;
1189 addr->addrtype = TIPC_ADDR_ID;
1190 memset(&addr->addr, 0, sizeof(addr->addr));
1191 addr->addr.id.ref = msg_origport(msg);
1192 addr->addr.id.node = msg_orignode(msg);
1193 addr->addr.name.domain = 0; /* could leave uninitialized */
1194 addr->scope = 0; /* could leave uninitialized */
1195 m->msg_namelen = sizeof(struct sockaddr_tipc);
1200 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1201 * @m: descriptor for message info
1202 * @msg: received message header
1203 * @tsk: TIPC port associated with message
1205 * Note: Ancillary data is not captured if not requested by receiver.
1207 * Returns 0 if successful, otherwise errno
1209 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1210 struct tipc_sock *tsk)
1218 if (likely(m->msg_controllen == 0))
1221 /* Optionally capture errored message object(s) */
1222 err = msg ? msg_errcode(msg) : 0;
1223 if (unlikely(err)) {
1225 anc_data[1] = msg_data_sz(msg);
1226 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1230 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1237 /* Optionally capture message destination object */
1238 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1239 switch (dest_type) {
1240 case TIPC_NAMED_MSG:
1242 anc_data[0] = msg_nametype(msg);
1243 anc_data[1] = msg_namelower(msg);
1244 anc_data[2] = msg_namelower(msg);
1246 case TIPC_MCAST_MSG:
1248 anc_data[0] = msg_nametype(msg);
1249 anc_data[1] = msg_namelower(msg);
1250 anc_data[2] = msg_nameupper(msg);
1253 has_name = (tsk->conn_type != 0);
1254 anc_data[0] = tsk->conn_type;
1255 anc_data[1] = tsk->conn_instance;
1256 anc_data[2] = tsk->conn_instance;
1262 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1270 static void tipc_sk_send_ack(struct tipc_sock *tsk, uint ack)
1272 struct net *net = sock_net(&tsk->sk);
1273 struct sk_buff *skb = NULL;
1274 struct tipc_msg *msg;
1275 u32 peer_port = tsk_peer_port(tsk);
1276 u32 dnode = tsk_peer_node(tsk);
1278 if (!tsk->connected)
1280 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1281 dnode, tsk_own_node(tsk), peer_port,
1282 tsk->portid, TIPC_OK);
1286 msg_set_msgcnt(msg, ack);
1287 tipc_link_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1290 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1292 struct sock *sk = sock->sk;
1294 long timeo = *timeop;
1298 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1299 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1300 if (sock->state == SS_DISCONNECTING) {
1305 timeo = schedule_timeout(timeo);
1309 if (!skb_queue_empty(&sk->sk_receive_queue))
1311 err = sock_intr_errno(timeo);
1312 if (signal_pending(current))
1318 finish_wait(sk_sleep(sk), &wait);
1324 * tipc_recvmsg - receive packet-oriented message
1326 * @m: descriptor for message info
1327 * @buf_len: total size of user buffer area
1328 * @flags: receive flags
1330 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1331 * If the complete message doesn't fit in user area, truncate it.
1333 * Returns size of returned message data, errno otherwise
1335 static int tipc_recvmsg(struct kiocb *iocb, struct socket *sock,
1336 struct msghdr *m, size_t buf_len, int flags)
1338 struct sock *sk = sock->sk;
1339 struct tipc_sock *tsk = tipc_sk(sk);
1340 struct sk_buff *buf;
1341 struct tipc_msg *msg;
1347 /* Catch invalid receive requests */
1348 if (unlikely(!buf_len))
1353 if (unlikely(sock->state == SS_UNCONNECTED)) {
1358 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1361 /* Look for a message in receive queue; wait if necessary */
1362 res = tipc_wait_for_rcvmsg(sock, &timeo);
1366 /* Look at first message in receive queue */
1367 buf = skb_peek(&sk->sk_receive_queue);
1369 sz = msg_data_sz(msg);
1370 err = msg_errcode(msg);
1372 /* Discard an empty non-errored message & try again */
1373 if ((!sz) && (!err)) {
1374 tsk_advance_rx_queue(sk);
1378 /* Capture sender's address (optional) */
1379 set_orig_addr(m, msg);
1381 /* Capture ancillary data (optional) */
1382 res = tipc_sk_anc_data_recv(m, msg, tsk);
1386 /* Capture message data (if valid) & compute return value (always) */
1388 if (unlikely(buf_len < sz)) {
1390 m->msg_flags |= MSG_TRUNC;
1392 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg), m, sz);
1397 if ((sock->state == SS_READY) ||
1398 ((err == TIPC_CONN_SHUTDOWN) || m->msg_control))
1404 /* Consume received message (optional) */
1405 if (likely(!(flags & MSG_PEEK))) {
1406 if ((sock->state != SS_READY) &&
1407 (++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1408 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1409 tsk->rcv_unacked = 0;
1411 tsk_advance_rx_queue(sk);
1419 * tipc_recv_stream - receive stream-oriented data
1421 * @m: descriptor for message info
1422 * @buf_len: total size of user buffer area
1423 * @flags: receive flags
1425 * Used for SOCK_STREAM messages only. If not enough data is available
1426 * will optionally wait for more; never truncates data.
1428 * Returns size of returned message data, errno otherwise
1430 static int tipc_recv_stream(struct kiocb *iocb, struct socket *sock,
1431 struct msghdr *m, size_t buf_len, int flags)
1433 struct sock *sk = sock->sk;
1434 struct tipc_sock *tsk = tipc_sk(sk);
1435 struct sk_buff *buf;
1436 struct tipc_msg *msg;
1439 int sz_to_copy, target, needed;
1444 /* Catch invalid receive attempts */
1445 if (unlikely(!buf_len))
1450 if (unlikely(sock->state == SS_UNCONNECTED)) {
1455 target = sock_rcvlowat(sk, flags & MSG_WAITALL, buf_len);
1456 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1459 /* Look for a message in receive queue; wait if necessary */
1460 res = tipc_wait_for_rcvmsg(sock, &timeo);
1464 /* Look at first message in receive queue */
1465 buf = skb_peek(&sk->sk_receive_queue);
1467 sz = msg_data_sz(msg);
1468 err = msg_errcode(msg);
1470 /* Discard an empty non-errored message & try again */
1471 if ((!sz) && (!err)) {
1472 tsk_advance_rx_queue(sk);
1476 /* Optionally capture sender's address & ancillary data of first msg */
1477 if (sz_copied == 0) {
1478 set_orig_addr(m, msg);
1479 res = tipc_sk_anc_data_recv(m, msg, tsk);
1484 /* Capture message data (if valid) & compute return value (always) */
1486 u32 offset = (u32)(unsigned long)(TIPC_SKB_CB(buf)->handle);
1489 needed = (buf_len - sz_copied);
1490 sz_to_copy = (sz <= needed) ? sz : needed;
1492 res = skb_copy_datagram_msg(buf, msg_hdr_sz(msg) + offset,
1497 sz_copied += sz_to_copy;
1499 if (sz_to_copy < sz) {
1500 if (!(flags & MSG_PEEK))
1501 TIPC_SKB_CB(buf)->handle =
1502 (void *)(unsigned long)(offset + sz_to_copy);
1507 goto exit; /* can't add error msg to valid data */
1509 if ((err == TIPC_CONN_SHUTDOWN) || m->msg_control)
1515 /* Consume received message (optional) */
1516 if (likely(!(flags & MSG_PEEK))) {
1517 if (unlikely(++tsk->rcv_unacked >= TIPC_CONNACK_INTV)) {
1518 tipc_sk_send_ack(tsk, tsk->rcv_unacked);
1519 tsk->rcv_unacked = 0;
1521 tsk_advance_rx_queue(sk);
1524 /* Loop around if more data is required */
1525 if ((sz_copied < buf_len) && /* didn't get all requested data */
1526 (!skb_queue_empty(&sk->sk_receive_queue) ||
1527 (sz_copied < target)) && /* and more is ready or required */
1528 (!(flags & MSG_PEEK)) && /* and aren't just peeking at data */
1529 (!err)) /* and haven't reached a FIN */
1534 return sz_copied ? sz_copied : res;
1538 * tipc_write_space - wake up thread if port congestion is released
1541 static void tipc_write_space(struct sock *sk)
1543 struct socket_wq *wq;
1546 wq = rcu_dereference(sk->sk_wq);
1547 if (wq_has_sleeper(wq))
1548 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
1549 POLLWRNORM | POLLWRBAND);
1554 * tipc_data_ready - wake up threads to indicate messages have been received
1556 * @len: the length of messages
1558 static void tipc_data_ready(struct sock *sk)
1560 struct socket_wq *wq;
1563 wq = rcu_dereference(sk->sk_wq);
1564 if (wq_has_sleeper(wq))
1565 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
1566 POLLRDNORM | POLLRDBAND);
1571 * filter_connect - Handle all incoming messages for a connection-based socket
1573 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1575 * Returns 0 (TIPC_OK) if everything ok, -TIPC_ERR_NO_PORT otherwise
1577 static int filter_connect(struct tipc_sock *tsk, struct sk_buff **skb)
1579 struct sock *sk = &tsk->sk;
1580 struct net *net = sock_net(sk);
1581 struct socket *sock = sk->sk_socket;
1582 struct tipc_msg *msg = buf_msg(*skb);
1583 int retval = -TIPC_ERR_NO_PORT;
1588 switch ((int)sock->state) {
1590 /* Accept only connection-based messages sent by peer */
1591 if (tsk_peer_msg(tsk, msg)) {
1592 if (unlikely(msg_errcode(msg))) {
1593 sock->state = SS_DISCONNECTING;
1595 /* let timer expire on it's own */
1596 tipc_node_remove_conn(net, tsk_peer_node(tsk),
1603 /* Accept only ACK or NACK message */
1605 if (unlikely(!msg_connected(msg)))
1608 if (unlikely(msg_errcode(msg))) {
1609 sock->state = SS_DISCONNECTING;
1610 sk->sk_err = ECONNREFUSED;
1615 if (unlikely(msg_importance(msg) > TIPC_CRITICAL_IMPORTANCE)) {
1616 sock->state = SS_DISCONNECTING;
1617 sk->sk_err = EINVAL;
1622 tipc_sk_finish_conn(tsk, msg_origport(msg), msg_orignode(msg));
1623 msg_set_importance(&tsk->phdr, msg_importance(msg));
1624 sock->state = SS_CONNECTED;
1626 /* If an incoming message is an 'ACK-', it should be
1627 * discarded here because it doesn't contain useful
1628 * data. In addition, we should try to wake up
1629 * connect() routine if sleeping.
1631 if (msg_data_sz(msg) == 0) {
1634 if (waitqueue_active(sk_sleep(sk)))
1635 wake_up_interruptible(sk_sleep(sk));
1640 case SS_UNCONNECTED:
1641 /* Accept only SYN message */
1642 if (!msg_connected(msg) && !(msg_errcode(msg)))
1645 case SS_DISCONNECTING:
1648 pr_err("Unknown socket state %u\n", sock->state);
1654 * rcvbuf_limit - get proper overload limit of socket receive queue
1658 * For all connection oriented messages, irrespective of importance,
1659 * the default overload value (i.e. 67MB) is set as limit.
1661 * For all connectionless messages, by default new queue limits are
1664 * TIPC_LOW_IMPORTANCE (4 MB)
1665 * TIPC_MEDIUM_IMPORTANCE (8 MB)
1666 * TIPC_HIGH_IMPORTANCE (16 MB)
1667 * TIPC_CRITICAL_IMPORTANCE (32 MB)
1669 * Returns overload limit according to corresponding message importance
1671 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *buf)
1673 struct tipc_msg *msg = buf_msg(buf);
1675 if (msg_connected(msg))
1676 return sysctl_tipc_rmem[2];
1678 return sk->sk_rcvbuf >> TIPC_CRITICAL_IMPORTANCE <<
1679 msg_importance(msg);
1683 * filter_rcv - validate incoming message
1685 * @skb: pointer to message. Set to NULL if buffer is consumed.
1687 * Enqueues message on receive queue if acceptable; optionally handles
1688 * disconnect indication for a connected socket.
1690 * Called with socket lock already taken
1692 * Returns 0 (TIPC_OK) if message was ok, -TIPC error code if rejected
1694 static int filter_rcv(struct sock *sk, struct sk_buff **skb)
1696 struct socket *sock = sk->sk_socket;
1697 struct tipc_sock *tsk = tipc_sk(sk);
1698 struct tipc_msg *msg = buf_msg(*skb);
1699 unsigned int limit = rcvbuf_limit(sk, *skb);
1702 if (unlikely(msg_user(msg) == CONN_MANAGER)) {
1703 tipc_sk_proto_rcv(tsk, skb);
1707 if (unlikely(msg_user(msg) == SOCK_WAKEUP)) {
1710 sk->sk_write_space(sk);
1715 /* Reject message if it is wrong sort of message for socket */
1716 if (msg_type(msg) > TIPC_DIRECT_MSG)
1717 return -TIPC_ERR_NO_PORT;
1719 if (sock->state == SS_READY) {
1720 if (msg_connected(msg))
1721 return -TIPC_ERR_NO_PORT;
1723 rc = filter_connect(tsk, skb);
1724 if (rc != TIPC_OK || !*skb)
1728 /* Reject message if there isn't room to queue it */
1729 if (sk_rmem_alloc_get(sk) + (*skb)->truesize >= limit)
1730 return -TIPC_ERR_OVERLOAD;
1732 /* Enqueue message */
1733 TIPC_SKB_CB(*skb)->handle = NULL;
1734 __skb_queue_tail(&sk->sk_receive_queue, *skb);
1735 skb_set_owner_r(*skb, sk);
1737 sk->sk_data_ready(sk);
1743 * tipc_backlog_rcv - handle incoming message from backlog queue
1747 * Caller must hold socket lock
1751 static int tipc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1756 struct tipc_sock *tsk = tipc_sk(sk);
1757 struct net *net = sock_net(sk);
1758 uint truesize = skb->truesize;
1760 err = filter_rcv(sk, &skb);
1762 dcnt = &tsk->dupl_rcvcnt;
1763 if (atomic_read(dcnt) < TIPC_CONN_OVERLOAD_LIMIT)
1764 atomic_add(truesize, dcnt);
1767 if (!err || tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode, -err))
1768 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
1773 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1774 * inputq and try adding them to socket or backlog queue
1775 * @inputq: list of incoming buffers with potentially different destinations
1776 * @sk: socket where the buffers should be enqueued
1777 * @dport: port number for the socket
1778 * @_skb: returned buffer to be forwarded or rejected, if applicable
1780 * Caller must hold socket lock
1782 * Returns TIPC_OK if all buffers enqueued, otherwise -TIPC_ERR_OVERLOAD
1783 * or -TIPC_ERR_NO_PORT
1785 static int tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
1786 u32 dport, struct sk_buff **_skb)
1791 struct sk_buff *skb;
1792 unsigned long time_limit = jiffies + 2;
1794 while (skb_queue_len(inputq)) {
1795 skb = tipc_skb_dequeue(inputq, dport);
1798 /* Return if softirq window exhausted */
1799 if (unlikely(time_after_eq(jiffies, time_limit)))
1801 if (!sock_owned_by_user(sk)) {
1802 err = filter_rcv(sk, &skb);
1808 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
1809 if (sk->sk_backlog.len)
1810 atomic_set(dcnt, 0);
1811 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
1812 if (likely(!sk_add_backlog(sk, skb, lim)))
1815 return -TIPC_ERR_OVERLOAD;
1821 * tipc_sk_rcv - handle a chain of incoming buffers
1822 * @inputq: buffer list containing the buffers
1823 * Consumes all buffers in list until inputq is empty
1824 * Note: may be called in multiple threads referring to the same queue
1825 * Returns 0 if last buffer was accepted, otherwise -EHOSTUNREACH
1826 * Only node local calls check the return value, sending single-buffer queues
1828 int tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
1830 u32 dnode, dport = 0;
1831 int err = -TIPC_ERR_NO_PORT;
1832 struct sk_buff *skb;
1833 struct tipc_sock *tsk;
1834 struct tipc_net *tn;
1837 while (skb_queue_len(inputq)) {
1839 dport = tipc_skb_peek_port(inputq, dport);
1840 tsk = tipc_sk_lookup(net, dport);
1843 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
1844 err = tipc_sk_enqueue(inputq, sk, dport, &skb);
1845 spin_unlock_bh(&sk->sk_lock.slock);
1850 skb = tipc_skb_dequeue(inputq, dport);
1854 if (tipc_msg_lookup_dest(net, skb, &dnode, &err))
1857 dnode = msg_destnode(buf_msg(skb));
1860 tn = net_generic(net, tipc_net_id);
1861 if (!tipc_msg_reverse(tn->own_addr, skb, &dnode, -err))
1864 tipc_link_xmit_skb(net, skb, dnode, dport);
1866 return err ? -EHOSTUNREACH : 0;
1869 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
1871 struct sock *sk = sock->sk;
1876 int err = sock_error(sk);
1881 if (signal_pending(current))
1882 return sock_intr_errno(*timeo_p);
1884 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1885 done = sk_wait_event(sk, timeo_p, sock->state != SS_CONNECTING);
1886 finish_wait(sk_sleep(sk), &wait);
1892 * tipc_connect - establish a connection to another TIPC port
1893 * @sock: socket structure
1894 * @dest: socket address for destination port
1895 * @destlen: size of socket address data structure
1896 * @flags: file-related flags associated with socket
1898 * Returns 0 on success, errno otherwise
1900 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
1901 int destlen, int flags)
1903 struct sock *sk = sock->sk;
1904 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
1905 struct msghdr m = {NULL,};
1906 long timeout = (flags & O_NONBLOCK) ? 0 : tipc_sk(sk)->conn_timeout;
1907 socket_state previous;
1912 /* For now, TIPC does not allow use of connect() with DGRAM/RDM types */
1913 if (sock->state == SS_READY) {
1919 * Reject connection attempt using multicast address
1921 * Note: send_msg() validates the rest of the address fields,
1922 * so there's no need to do it here
1924 if (dst->addrtype == TIPC_ADDR_MCAST) {
1929 previous = sock->state;
1930 switch (sock->state) {
1931 case SS_UNCONNECTED:
1932 /* Send a 'SYN-' to destination */
1934 m.msg_namelen = destlen;
1936 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1937 * indicate send_msg() is never blocked.
1940 m.msg_flags = MSG_DONTWAIT;
1942 res = tipc_sendmsg(NULL, sock, &m, 0);
1943 if ((res < 0) && (res != -EWOULDBLOCK))
1946 /* Just entered SS_CONNECTING state; the only
1947 * difference is that return value in non-blocking
1948 * case is EINPROGRESS, rather than EALREADY.
1952 if (previous == SS_CONNECTING)
1956 timeout = msecs_to_jiffies(timeout);
1957 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1958 res = tipc_wait_for_connect(sock, &timeout);
1973 * tipc_listen - allow socket to listen for incoming connections
1974 * @sock: socket structure
1977 * Returns 0 on success, errno otherwise
1979 static int tipc_listen(struct socket *sock, int len)
1981 struct sock *sk = sock->sk;
1986 if (sock->state != SS_UNCONNECTED)
1989 sock->state = SS_LISTENING;
1997 static int tipc_wait_for_accept(struct socket *sock, long timeo)
1999 struct sock *sk = sock->sk;
2003 /* True wake-one mechanism for incoming connections: only
2004 * one process gets woken up, not the 'whole herd'.
2005 * Since we do not 'race & poll' for established sockets
2006 * anymore, the common case will execute the loop only once.
2009 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2010 TASK_INTERRUPTIBLE);
2011 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2013 timeo = schedule_timeout(timeo);
2017 if (!skb_queue_empty(&sk->sk_receive_queue))
2020 if (sock->state != SS_LISTENING)
2022 err = sock_intr_errno(timeo);
2023 if (signal_pending(current))
2029 finish_wait(sk_sleep(sk), &wait);
2034 * tipc_accept - wait for connection request
2035 * @sock: listening socket
2036 * @newsock: new socket that is to be connected
2037 * @flags: file-related flags associated with socket
2039 * Returns 0 on success, errno otherwise
2041 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags)
2043 struct sock *new_sk, *sk = sock->sk;
2044 struct sk_buff *buf;
2045 struct tipc_sock *new_tsock;
2046 struct tipc_msg *msg;
2052 if (sock->state != SS_LISTENING) {
2056 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2057 res = tipc_wait_for_accept(sock, timeo);
2061 buf = skb_peek(&sk->sk_receive_queue);
2063 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, 1);
2067 new_sk = new_sock->sk;
2068 new_tsock = tipc_sk(new_sk);
2071 /* we lock on new_sk; but lockdep sees the lock on sk */
2072 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2075 * Reject any stray messages received by new socket
2076 * before the socket lock was taken (very, very unlikely)
2078 tsk_rej_rx_queue(new_sk);
2080 /* Connect new socket to it's peer */
2081 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2082 new_sock->state = SS_CONNECTED;
2084 tsk_set_importance(new_tsock, msg_importance(msg));
2085 if (msg_named(msg)) {
2086 new_tsock->conn_type = msg_nametype(msg);
2087 new_tsock->conn_instance = msg_nameinst(msg);
2091 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2092 * Respond to 'SYN+' by queuing it on new socket.
2094 if (!msg_data_sz(msg)) {
2095 struct msghdr m = {NULL,};
2097 tsk_advance_rx_queue(sk);
2098 tipc_send_packet(NULL, new_sock, &m, 0);
2100 __skb_dequeue(&sk->sk_receive_queue);
2101 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2102 skb_set_owner_r(buf, new_sk);
2104 release_sock(new_sk);
2111 * tipc_shutdown - shutdown socket connection
2112 * @sock: socket structure
2113 * @how: direction to close (must be SHUT_RDWR)
2115 * Terminates connection (if necessary), then purges socket's receive queue.
2117 * Returns 0 on success, errno otherwise
2119 static int tipc_shutdown(struct socket *sock, int how)
2121 struct sock *sk = sock->sk;
2122 struct net *net = sock_net(sk);
2123 struct tipc_sock *tsk = tipc_sk(sk);
2124 struct sk_buff *skb;
2128 if (how != SHUT_RDWR)
2133 switch (sock->state) {
2138 /* Disconnect and send a 'FIN+' or 'FIN-' message to peer */
2139 skb = __skb_dequeue(&sk->sk_receive_queue);
2141 if (TIPC_SKB_CB(skb)->handle != NULL) {
2145 if (tipc_msg_reverse(tsk_own_node(tsk), skb, &dnode,
2146 TIPC_CONN_SHUTDOWN))
2147 tipc_link_xmit_skb(net, skb, dnode,
2149 tipc_node_remove_conn(net, dnode, tsk->portid);
2151 dnode = tsk_peer_node(tsk);
2153 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2154 TIPC_CONN_MSG, SHORT_H_SIZE,
2155 0, dnode, tsk_own_node(tsk),
2157 tsk->portid, TIPC_CONN_SHUTDOWN);
2158 tipc_link_xmit_skb(net, skb, dnode, tsk->portid);
2161 sock->state = SS_DISCONNECTING;
2162 tipc_node_remove_conn(net, dnode, tsk->portid);
2165 case SS_DISCONNECTING:
2167 /* Discard any unreceived messages */
2168 __skb_queue_purge(&sk->sk_receive_queue);
2170 /* Wake up anyone sleeping in poll */
2171 sk->sk_state_change(sk);
2183 static void tipc_sk_timeout(unsigned long data)
2185 struct tipc_sock *tsk = (struct tipc_sock *)data;
2186 struct sock *sk = &tsk->sk;
2187 struct sk_buff *skb = NULL;
2188 u32 peer_port, peer_node;
2189 u32 own_node = tsk_own_node(tsk);
2192 if (!tsk->connected) {
2196 peer_port = tsk_peer_port(tsk);
2197 peer_node = tsk_peer_node(tsk);
2199 if (tsk->probing_state == TIPC_CONN_PROBING) {
2200 /* Previous probe not answered -> self abort */
2201 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
2202 TIPC_CONN_MSG, SHORT_H_SIZE, 0,
2203 own_node, peer_node, tsk->portid,
2204 peer_port, TIPC_ERR_NO_PORT);
2206 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE,
2207 INT_H_SIZE, 0, peer_node, own_node,
2208 peer_port, tsk->portid, TIPC_OK);
2209 tsk->probing_state = TIPC_CONN_PROBING;
2210 sk_reset_timer(sk, &sk->sk_timer, jiffies + tsk->probing_intv);
2214 tipc_link_xmit_skb(sock_net(sk), skb, peer_node, tsk->portid);
2219 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2220 struct tipc_name_seq const *seq)
2222 struct net *net = sock_net(&tsk->sk);
2223 struct publication *publ;
2228 key = tsk->portid + tsk->pub_count + 1;
2229 if (key == tsk->portid)
2232 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2233 scope, tsk->portid, key);
2234 if (unlikely(!publ))
2237 list_add(&publ->pport_list, &tsk->publications);
2243 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2244 struct tipc_name_seq const *seq)
2246 struct net *net = sock_net(&tsk->sk);
2247 struct publication *publ;
2248 struct publication *safe;
2251 list_for_each_entry_safe(publ, safe, &tsk->publications, pport_list) {
2253 if (publ->scope != scope)
2255 if (publ->type != seq->type)
2257 if (publ->lower != seq->lower)
2259 if (publ->upper != seq->upper)
2261 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2262 publ->ref, publ->key);
2266 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2267 publ->ref, publ->key);
2270 if (list_empty(&tsk->publications))
2275 static int tipc_sk_show(struct tipc_sock *tsk, char *buf,
2276 int len, int full_id)
2278 struct net *net = sock_net(&tsk->sk);
2279 struct tipc_net *tn = net_generic(net, tipc_net_id);
2280 struct publication *publ;
2284 ret = tipc_snprintf(buf, len, "<%u.%u.%u:%u>:",
2285 tipc_zone(tn->own_addr),
2286 tipc_cluster(tn->own_addr),
2287 tipc_node(tn->own_addr), tsk->portid);
2289 ret = tipc_snprintf(buf, len, "%-10u:", tsk->portid);
2291 if (tsk->connected) {
2292 u32 dport = tsk_peer_port(tsk);
2293 u32 destnode = tsk_peer_node(tsk);
2295 ret += tipc_snprintf(buf + ret, len - ret,
2296 " connected to <%u.%u.%u:%u>",
2297 tipc_zone(destnode),
2298 tipc_cluster(destnode),
2299 tipc_node(destnode), dport);
2300 if (tsk->conn_type != 0)
2301 ret += tipc_snprintf(buf + ret, len - ret,
2302 " via {%u,%u}", tsk->conn_type,
2303 tsk->conn_instance);
2304 } else if (tsk->published) {
2305 ret += tipc_snprintf(buf + ret, len - ret, " bound to");
2306 list_for_each_entry(publ, &tsk->publications, pport_list) {
2307 if (publ->lower == publ->upper)
2308 ret += tipc_snprintf(buf + ret, len - ret,
2309 " {%u,%u}", publ->type,
2312 ret += tipc_snprintf(buf + ret, len - ret,
2313 " {%u,%u,%u}", publ->type,
2314 publ->lower, publ->upper);
2317 ret += tipc_snprintf(buf + ret, len - ret, "\n");
2321 struct sk_buff *tipc_sk_socks_show(struct net *net)
2323 struct tipc_net *tn = net_generic(net, tipc_net_id);
2324 const struct bucket_table *tbl;
2325 struct rhash_head *pos;
2326 struct sk_buff *buf;
2327 struct tlv_desc *rep_tlv;
2330 struct tipc_sock *tsk;
2334 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2337 rep_tlv = (struct tlv_desc *)buf->data;
2338 pb = TLV_DATA(rep_tlv);
2339 pb_len = ULTRA_STRING_MAX_LEN;
2342 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2343 for (i = 0; i < tbl->size; i++) {
2344 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2345 spin_lock_bh(&tsk->sk.sk_lock.slock);
2346 str_len += tipc_sk_show(tsk, pb + str_len,
2347 pb_len - str_len, 0);
2348 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2353 str_len += 1; /* for "\0" */
2354 skb_put(buf, TLV_SPACE(str_len));
2355 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2360 /* tipc_sk_reinit: set non-zero address in all existing sockets
2361 * when we go from standalone to network mode.
2363 void tipc_sk_reinit(struct net *net)
2365 struct tipc_net *tn = net_generic(net, tipc_net_id);
2366 const struct bucket_table *tbl;
2367 struct rhash_head *pos;
2368 struct tipc_sock *tsk;
2369 struct tipc_msg *msg;
2373 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2374 for (i = 0; i < tbl->size; i++) {
2375 rht_for_each_entry_rcu(tsk, pos, tbl, i, node) {
2376 spin_lock_bh(&tsk->sk.sk_lock.slock);
2378 msg_set_prevnode(msg, tn->own_addr);
2379 msg_set_orignode(msg, tn->own_addr);
2380 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2386 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2388 struct tipc_net *tn = net_generic(net, tipc_net_id);
2389 struct tipc_sock *tsk;
2392 tsk = rhashtable_lookup(&tn->sk_rht, &portid);
2394 sock_hold(&tsk->sk);
2400 static int tipc_sk_insert(struct tipc_sock *tsk)
2402 struct sock *sk = &tsk->sk;
2403 struct net *net = sock_net(sk);
2404 struct tipc_net *tn = net_generic(net, tipc_net_id);
2405 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2406 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2408 while (remaining--) {
2410 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2411 portid = TIPC_MIN_PORT;
2412 tsk->portid = portid;
2413 sock_hold(&tsk->sk);
2414 if (rhashtable_lookup_insert(&tn->sk_rht, &tsk->node))
2422 static void tipc_sk_remove(struct tipc_sock *tsk)
2424 struct sock *sk = &tsk->sk;
2425 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2427 if (rhashtable_remove(&tn->sk_rht, &tsk->node)) {
2428 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
2433 int tipc_sk_rht_init(struct net *net)
2435 struct tipc_net *tn = net_generic(net, tipc_net_id);
2436 struct rhashtable_params rht_params = {
2438 .head_offset = offsetof(struct tipc_sock, node),
2439 .key_offset = offsetof(struct tipc_sock, portid),
2440 .key_len = sizeof(u32), /* portid */
2442 .max_shift = 20, /* 1M */
2443 .min_shift = 8, /* 256 */
2444 .grow_decision = rht_grow_above_75,
2445 .shrink_decision = rht_shrink_below_30,
2448 return rhashtable_init(&tn->sk_rht, &rht_params);
2451 void tipc_sk_rht_destroy(struct net *net)
2453 struct tipc_net *tn = net_generic(net, tipc_net_id);
2455 /* Wait for socket readers to complete */
2458 rhashtable_destroy(&tn->sk_rht);
2462 * tipc_setsockopt - set socket option
2463 * @sock: socket structure
2464 * @lvl: option level
2465 * @opt: option identifier
2466 * @ov: pointer to new option value
2467 * @ol: length of option value
2469 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2470 * (to ease compatibility).
2472 * Returns 0 on success, errno otherwise
2474 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2475 char __user *ov, unsigned int ol)
2477 struct sock *sk = sock->sk;
2478 struct tipc_sock *tsk = tipc_sk(sk);
2482 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2484 if (lvl != SOL_TIPC)
2485 return -ENOPROTOOPT;
2486 if (ol < sizeof(value))
2488 res = get_user(value, (u32 __user *)ov);
2495 case TIPC_IMPORTANCE:
2496 res = tsk_set_importance(tsk, value);
2498 case TIPC_SRC_DROPPABLE:
2499 if (sock->type != SOCK_STREAM)
2500 tsk_set_unreliable(tsk, value);
2504 case TIPC_DEST_DROPPABLE:
2505 tsk_set_unreturnable(tsk, value);
2507 case TIPC_CONN_TIMEOUT:
2508 tipc_sk(sk)->conn_timeout = value;
2509 /* no need to set "res", since already 0 at this point */
2521 * tipc_getsockopt - get socket option
2522 * @sock: socket structure
2523 * @lvl: option level
2524 * @opt: option identifier
2525 * @ov: receptacle for option value
2526 * @ol: receptacle for length of option value
2528 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2529 * (to ease compatibility).
2531 * Returns 0 on success, errno otherwise
2533 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2534 char __user *ov, int __user *ol)
2536 struct sock *sk = sock->sk;
2537 struct tipc_sock *tsk = tipc_sk(sk);
2542 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2543 return put_user(0, ol);
2544 if (lvl != SOL_TIPC)
2545 return -ENOPROTOOPT;
2546 res = get_user(len, ol);
2553 case TIPC_IMPORTANCE:
2554 value = tsk_importance(tsk);
2556 case TIPC_SRC_DROPPABLE:
2557 value = tsk_unreliable(tsk);
2559 case TIPC_DEST_DROPPABLE:
2560 value = tsk_unreturnable(tsk);
2562 case TIPC_CONN_TIMEOUT:
2563 value = tsk->conn_timeout;
2564 /* no need to set "res", since already 0 at this point */
2566 case TIPC_NODE_RECVQ_DEPTH:
2567 value = 0; /* was tipc_queue_size, now obsolete */
2569 case TIPC_SOCK_RECVQ_DEPTH:
2570 value = skb_queue_len(&sk->sk_receive_queue);
2579 return res; /* "get" failed */
2581 if (len < sizeof(value))
2584 if (copy_to_user(ov, &value, sizeof(value)))
2587 return put_user(sizeof(value), ol);
2590 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2592 struct sock *sk = sock->sk;
2593 struct tipc_sioc_ln_req lnr;
2594 void __user *argp = (void __user *)arg;
2597 case SIOCGETLINKNAME:
2598 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2600 if (!tipc_node_get_linkname(sock_net(sk),
2601 lnr.bearer_id & 0xffff, lnr.peer,
2602 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2603 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2607 return -EADDRNOTAVAIL;
2609 return -ENOIOCTLCMD;
2613 /* Protocol switches for the various types of TIPC sockets */
2615 static const struct proto_ops msg_ops = {
2616 .owner = THIS_MODULE,
2618 .release = tipc_release,
2620 .connect = tipc_connect,
2621 .socketpair = sock_no_socketpair,
2622 .accept = sock_no_accept,
2623 .getname = tipc_getname,
2625 .ioctl = tipc_ioctl,
2626 .listen = sock_no_listen,
2627 .shutdown = tipc_shutdown,
2628 .setsockopt = tipc_setsockopt,
2629 .getsockopt = tipc_getsockopt,
2630 .sendmsg = tipc_sendmsg,
2631 .recvmsg = tipc_recvmsg,
2632 .mmap = sock_no_mmap,
2633 .sendpage = sock_no_sendpage
2636 static const struct proto_ops packet_ops = {
2637 .owner = THIS_MODULE,
2639 .release = tipc_release,
2641 .connect = tipc_connect,
2642 .socketpair = sock_no_socketpair,
2643 .accept = tipc_accept,
2644 .getname = tipc_getname,
2646 .ioctl = tipc_ioctl,
2647 .listen = tipc_listen,
2648 .shutdown = tipc_shutdown,
2649 .setsockopt = tipc_setsockopt,
2650 .getsockopt = tipc_getsockopt,
2651 .sendmsg = tipc_send_packet,
2652 .recvmsg = tipc_recvmsg,
2653 .mmap = sock_no_mmap,
2654 .sendpage = sock_no_sendpage
2657 static const struct proto_ops stream_ops = {
2658 .owner = THIS_MODULE,
2660 .release = tipc_release,
2662 .connect = tipc_connect,
2663 .socketpair = sock_no_socketpair,
2664 .accept = tipc_accept,
2665 .getname = tipc_getname,
2667 .ioctl = tipc_ioctl,
2668 .listen = tipc_listen,
2669 .shutdown = tipc_shutdown,
2670 .setsockopt = tipc_setsockopt,
2671 .getsockopt = tipc_getsockopt,
2672 .sendmsg = tipc_send_stream,
2673 .recvmsg = tipc_recv_stream,
2674 .mmap = sock_no_mmap,
2675 .sendpage = sock_no_sendpage
2678 static const struct net_proto_family tipc_family_ops = {
2679 .owner = THIS_MODULE,
2681 .create = tipc_sk_create
2684 static struct proto tipc_proto = {
2686 .owner = THIS_MODULE,
2687 .obj_size = sizeof(struct tipc_sock),
2688 .sysctl_rmem = sysctl_tipc_rmem
2691 static struct proto tipc_proto_kern = {
2693 .obj_size = sizeof(struct tipc_sock),
2694 .sysctl_rmem = sysctl_tipc_rmem
2698 * tipc_socket_init - initialize TIPC socket interface
2700 * Returns 0 on success, errno otherwise
2702 int tipc_socket_init(void)
2706 res = proto_register(&tipc_proto, 1);
2708 pr_err("Failed to register TIPC protocol type\n");
2712 res = sock_register(&tipc_family_ops);
2714 pr_err("Failed to register TIPC socket type\n");
2715 proto_unregister(&tipc_proto);
2723 * tipc_socket_stop - stop TIPC socket interface
2725 void tipc_socket_stop(void)
2727 sock_unregister(tipc_family_ops.family);
2728 proto_unregister(&tipc_proto);
2731 /* Caller should hold socket lock for the passed tipc socket. */
2732 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
2736 struct nlattr *nest;
2738 peer_node = tsk_peer_node(tsk);
2739 peer_port = tsk_peer_port(tsk);
2741 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
2743 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
2745 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
2748 if (tsk->conn_type != 0) {
2749 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
2751 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
2753 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
2756 nla_nest_end(skb, nest);
2761 nla_nest_cancel(skb, nest);
2766 /* Caller should hold socket lock for the passed tipc socket. */
2767 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
2768 struct tipc_sock *tsk)
2772 struct nlattr *attrs;
2773 struct net *net = sock_net(skb->sk);
2774 struct tipc_net *tn = net_generic(net, tipc_net_id);
2776 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2777 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
2781 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
2783 goto genlmsg_cancel;
2784 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid))
2785 goto attr_msg_cancel;
2786 if (nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tn->own_addr))
2787 goto attr_msg_cancel;
2789 if (tsk->connected) {
2790 err = __tipc_nl_add_sk_con(skb, tsk);
2792 goto attr_msg_cancel;
2793 } else if (!list_empty(&tsk->publications)) {
2794 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
2795 goto attr_msg_cancel;
2797 nla_nest_end(skb, attrs);
2798 genlmsg_end(skb, hdr);
2803 nla_nest_cancel(skb, attrs);
2805 genlmsg_cancel(skb, hdr);
2810 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
2813 struct tipc_sock *tsk;
2814 const struct bucket_table *tbl;
2815 struct rhash_head *pos;
2816 struct net *net = sock_net(skb->sk);
2817 struct tipc_net *tn = net_generic(net, tipc_net_id);
2818 u32 tbl_id = cb->args[0];
2819 u32 prev_portid = cb->args[1];
2822 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
2823 for (; tbl_id < tbl->size; tbl_id++) {
2824 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
2825 spin_lock_bh(&tsk->sk.sk_lock.slock);
2826 if (prev_portid && prev_portid != tsk->portid) {
2827 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2831 err = __tipc_nl_add_sk(skb, cb, tsk);
2833 prev_portid = tsk->portid;
2834 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2838 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2843 cb->args[0] = tbl_id;
2844 cb->args[1] = prev_portid;
2849 /* Caller should hold socket lock for the passed tipc socket. */
2850 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
2851 struct netlink_callback *cb,
2852 struct publication *publ)
2855 struct nlattr *attrs;
2857 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2858 &tipc_genl_v2_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
2862 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
2864 goto genlmsg_cancel;
2866 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
2867 goto attr_msg_cancel;
2868 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
2869 goto attr_msg_cancel;
2870 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
2871 goto attr_msg_cancel;
2872 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
2873 goto attr_msg_cancel;
2875 nla_nest_end(skb, attrs);
2876 genlmsg_end(skb, hdr);
2881 nla_nest_cancel(skb, attrs);
2883 genlmsg_cancel(skb, hdr);
2888 /* Caller should hold socket lock for the passed tipc socket. */
2889 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
2890 struct netlink_callback *cb,
2891 struct tipc_sock *tsk, u32 *last_publ)
2894 struct publication *p;
2897 list_for_each_entry(p, &tsk->publications, pport_list) {
2898 if (p->key == *last_publ)
2901 if (p->key != *last_publ) {
2902 /* We never set seq or call nl_dump_check_consistent()
2903 * this means that setting prev_seq here will cause the
2904 * consistence check to fail in the netlink callback
2905 * handler. Resulting in the last NLMSG_DONE message
2906 * having the NLM_F_DUMP_INTR flag set.
2913 p = list_first_entry(&tsk->publications, struct publication,
2917 list_for_each_entry_from(p, &tsk->publications, pport_list) {
2918 err = __tipc_nl_add_sk_publ(skb, cb, p);
2920 *last_publ = p->key;
2929 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
2932 u32 tsk_portid = cb->args[0];
2933 u32 last_publ = cb->args[1];
2934 u32 done = cb->args[2];
2935 struct net *net = sock_net(skb->sk);
2936 struct tipc_sock *tsk;
2939 struct nlattr **attrs;
2940 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
2942 err = tipc_nlmsg_parse(cb->nlh, &attrs);
2946 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
2947 attrs[TIPC_NLA_SOCK],
2948 tipc_nl_sock_policy);
2952 if (!sock[TIPC_NLA_SOCK_REF])
2955 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
2961 tsk = tipc_sk_lookup(net, tsk_portid);
2965 lock_sock(&tsk->sk);
2966 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
2969 release_sock(&tsk->sk);
2972 cb->args[0] = tsk_portid;
2973 cb->args[1] = last_publ;