2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2017, 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/sched/signal.h>
41 #include "name_table.h"
44 #include "name_distr.h"
50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
52 #define TIPC_FWD_MSG 1
53 #define TIPC_MAX_PORT 0xffffffff
54 #define TIPC_MIN_PORT 1
55 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
58 TIPC_LISTEN = TCP_LISTEN,
59 TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 TIPC_OPEN = TCP_CLOSE,
61 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 TIPC_CONNECTING = TCP_SYN_SENT,
65 struct sockaddr_pair {
66 struct sockaddr_tipc sock;
67 struct sockaddr_tipc member;
71 * struct tipc_sock - TIPC socket structure
72 * @sk: socket - interacts with 'port' and with user via the socket API
73 * @conn_type: TIPC type used when connection was established
74 * @conn_instance: TIPC instance used when connection was established
75 * @published: non-zero if port has one or more associated names
76 * @max_pkt: maximum packet size "hint" used when building messages sent by port
77 * @portid: unique port identity in TIPC socket hash table
78 * @phdr: preformatted message header used when sending messages
79 * #cong_links: list of congested links
80 * @publications: list of publications for port
81 * @blocking_link: address of the congested link we are currently sleeping on
82 * @pub_count: total # of publications port has made during its lifetime
83 * @conn_timeout: the time we can wait for an unresponded setup request
84 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
85 * @cong_link_cnt: number of congested links
86 * @snt_unacked: # messages sent by socket, and not yet acked by peer
87 * @rcv_unacked: # messages read by user, but not yet acked back to peer
88 * @peer: 'connected' peer for dgram/rdm
89 * @node: hash table node
90 * @mc_method: cookie for use between socket and broadcast layer
91 * @rcu: rcu struct for tipc_sock
100 struct tipc_msg phdr;
101 struct list_head cong_links;
102 struct list_head publications;
104 atomic_t dupl_rcvcnt;
113 struct sockaddr_tipc peer;
114 struct rhash_head node;
115 struct tipc_mc_method mc_method;
117 struct tipc_group *group;
121 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
122 static void tipc_data_ready(struct sock *sk);
123 static void tipc_write_space(struct sock *sk);
124 static void tipc_sock_destruct(struct sock *sk);
125 static int tipc_release(struct socket *sock);
126 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
128 static void tipc_sk_timeout(struct timer_list *t);
129 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
130 struct tipc_name_seq const *seq);
131 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
132 struct tipc_name_seq const *seq);
133 static int tipc_sk_leave(struct tipc_sock *tsk);
134 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
135 static int tipc_sk_insert(struct tipc_sock *tsk);
136 static void tipc_sk_remove(struct tipc_sock *tsk);
137 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
138 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
140 static const struct proto_ops packet_ops;
141 static const struct proto_ops stream_ops;
142 static const struct proto_ops msg_ops;
143 static struct proto tipc_proto;
144 static const struct rhashtable_params tsk_rht_params;
146 static u32 tsk_own_node(struct tipc_sock *tsk)
148 return msg_prevnode(&tsk->phdr);
151 static u32 tsk_peer_node(struct tipc_sock *tsk)
153 return msg_destnode(&tsk->phdr);
156 static u32 tsk_peer_port(struct tipc_sock *tsk)
158 return msg_destport(&tsk->phdr);
161 static bool tsk_unreliable(struct tipc_sock *tsk)
163 return msg_src_droppable(&tsk->phdr) != 0;
166 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
168 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
171 static bool tsk_unreturnable(struct tipc_sock *tsk)
173 return msg_dest_droppable(&tsk->phdr) != 0;
176 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
178 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
181 static int tsk_importance(struct tipc_sock *tsk)
183 return msg_importance(&tsk->phdr);
186 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
188 if (imp > TIPC_CRITICAL_IMPORTANCE)
190 msg_set_importance(&tsk->phdr, (u32)imp);
194 static struct tipc_sock *tipc_sk(const struct sock *sk)
196 return container_of(sk, struct tipc_sock, sk);
199 static bool tsk_conn_cong(struct tipc_sock *tsk)
201 return tsk->snt_unacked > tsk->snd_win;
204 static u16 tsk_blocks(int len)
206 return ((len / FLOWCTL_BLK_SZ) + 1);
209 /* tsk_blocks(): translate a buffer size in bytes to number of
210 * advertisable blocks, taking into account the ratio truesize(len)/len
211 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
213 static u16 tsk_adv_blocks(int len)
215 return len / FLOWCTL_BLK_SZ / 4;
218 /* tsk_inc(): increment counter for sent or received data
219 * - If block based flow control is not supported by peer we
220 * fall back to message based ditto, incrementing the counter
222 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
224 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
225 return ((msglen / FLOWCTL_BLK_SZ) + 1);
230 * tsk_advance_rx_queue - discard first buffer in socket receive queue
232 * Caller must hold socket lock
234 static void tsk_advance_rx_queue(struct sock *sk)
236 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
239 /* tipc_sk_respond() : send response message back to sender
241 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
245 u32 onode = tipc_own_addr(sock_net(sk));
247 if (!tipc_msg_reverse(onode, &skb, err))
250 dnode = msg_destnode(buf_msg(skb));
251 selector = msg_origport(buf_msg(skb));
252 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
256 * tsk_rej_rx_queue - reject all buffers in socket receive queue
258 * Caller must hold socket lock
260 static void tsk_rej_rx_queue(struct sock *sk)
264 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
265 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
268 static bool tipc_sk_connected(struct sock *sk)
270 return sk->sk_state == TIPC_ESTABLISHED;
273 /* tipc_sk_type_connectionless - check if the socket is datagram socket
276 * Returns true if connection less, false otherwise
278 static bool tipc_sk_type_connectionless(struct sock *sk)
280 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
283 /* tsk_peer_msg - verify if message was sent by connected port's peer
285 * Handles cases where the node's network address has changed from
286 * the default of <0.0.0> to its configured setting.
288 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
290 struct sock *sk = &tsk->sk;
291 u32 self = tipc_own_addr(sock_net(sk));
292 u32 peer_port = tsk_peer_port(tsk);
293 u32 orig_node, peer_node;
295 if (unlikely(!tipc_sk_connected(sk)))
298 if (unlikely(msg_origport(msg) != peer_port))
301 orig_node = msg_orignode(msg);
302 peer_node = tsk_peer_node(tsk);
304 if (likely(orig_node == peer_node))
307 if (!orig_node && peer_node == self)
310 if (!peer_node && orig_node == self)
316 /* tipc_set_sk_state - set the sk_state of the socket
319 * Caller must hold socket lock
321 * Returns 0 on success, errno otherwise
323 static int tipc_set_sk_state(struct sock *sk, int state)
325 int oldsk_state = sk->sk_state;
333 case TIPC_CONNECTING:
334 if (oldsk_state == TIPC_OPEN)
337 case TIPC_ESTABLISHED:
338 if (oldsk_state == TIPC_CONNECTING ||
339 oldsk_state == TIPC_OPEN)
342 case TIPC_DISCONNECTING:
343 if (oldsk_state == TIPC_CONNECTING ||
344 oldsk_state == TIPC_ESTABLISHED)
350 sk->sk_state = state;
355 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
357 struct sock *sk = sock->sk;
358 int err = sock_error(sk);
359 int typ = sock->type;
363 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
364 if (sk->sk_state == TIPC_DISCONNECTING)
366 else if (!tipc_sk_connected(sk))
371 if (signal_pending(current))
372 return sock_intr_errno(*timeout);
377 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
382 while ((rc_ = !(condition_))) { \
383 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
385 rc_ = tipc_sk_sock_err((sock_), timeo_); \
388 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \
390 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
391 sched_annotate_sleep(); \
393 remove_wait_queue(sk_sleep(sk_), &wait_); \
399 * tipc_sk_create - create a TIPC socket
400 * @net: network namespace (must be default network)
401 * @sock: pre-allocated socket structure
402 * @protocol: protocol indicator (must be 0)
403 * @kern: caused by kernel or by userspace?
405 * This routine creates additional data structures used by the TIPC socket,
406 * initializes them, and links them together.
408 * Returns 0 on success, errno otherwise
410 static int tipc_sk_create(struct net *net, struct socket *sock,
411 int protocol, int kern)
413 const struct proto_ops *ops;
415 struct tipc_sock *tsk;
416 struct tipc_msg *msg;
418 /* Validate arguments */
419 if (unlikely(protocol != 0))
420 return -EPROTONOSUPPORT;
422 switch (sock->type) {
437 /* Allocate socket's protocol area */
438 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
443 tsk->max_pkt = MAX_PKT_DEFAULT;
444 INIT_LIST_HEAD(&tsk->publications);
445 INIT_LIST_HEAD(&tsk->cong_links);
448 /* Finish initializing socket data structures */
450 sock_init_data(sock, sk);
451 tipc_set_sk_state(sk, TIPC_OPEN);
452 if (tipc_sk_insert(tsk)) {
453 pr_warn("Socket create failed; port number exhausted\n");
457 /* Ensure tsk is visible before we read own_addr. */
460 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
461 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
463 msg_set_origport(msg, tsk->portid);
464 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
466 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
467 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
468 sk->sk_data_ready = tipc_data_ready;
469 sk->sk_write_space = tipc_write_space;
470 sk->sk_destruct = tipc_sock_destruct;
471 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
472 tsk->group_is_open = true;
473 atomic_set(&tsk->dupl_rcvcnt, 0);
475 /* Start out with safe limits until we receive an advertised window */
476 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
477 tsk->rcv_win = tsk->snd_win;
479 if (tipc_sk_type_connectionless(sk)) {
480 tsk_set_unreturnable(tsk, true);
481 if (sock->type == SOCK_DGRAM)
482 tsk_set_unreliable(tsk, true);
488 static void tipc_sk_callback(struct rcu_head *head)
490 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
495 /* Caller should hold socket lock for the socket. */
496 static void __tipc_shutdown(struct socket *sock, int error)
498 struct sock *sk = sock->sk;
499 struct tipc_sock *tsk = tipc_sk(sk);
500 struct net *net = sock_net(sk);
501 long timeout = CONN_TIMEOUT_DEFAULT;
502 u32 dnode = tsk_peer_node(tsk);
505 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
506 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
507 !tsk_conn_cong(tsk)));
509 /* Remove any pending SYN message */
510 __skb_queue_purge(&sk->sk_write_queue);
512 /* Reject all unreceived messages, except on an active connection
513 * (which disconnects locally & sends a 'FIN+' to peer).
515 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
516 if (TIPC_SKB_CB(skb)->bytes_read) {
520 if (!tipc_sk_type_connectionless(sk) &&
521 sk->sk_state != TIPC_DISCONNECTING) {
522 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
523 tipc_node_remove_conn(net, dnode, tsk->portid);
525 tipc_sk_respond(sk, skb, error);
528 if (tipc_sk_type_connectionless(sk))
531 if (sk->sk_state != TIPC_DISCONNECTING) {
532 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
533 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
534 tsk_own_node(tsk), tsk_peer_port(tsk),
537 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
538 tipc_node_remove_conn(net, dnode, tsk->portid);
539 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
544 * tipc_release - destroy a TIPC socket
545 * @sock: socket to destroy
547 * This routine cleans up any messages that are still queued on the socket.
548 * For DGRAM and RDM socket types, all queued messages are rejected.
549 * For SEQPACKET and STREAM socket types, the first message is rejected
550 * and any others are discarded. (If the first message on a STREAM socket
551 * is partially-read, it is discarded and the next one is rejected instead.)
553 * NOTE: Rejected messages are not necessarily returned to the sender! They
554 * are returned or discarded according to the "destination droppable" setting
555 * specified for the message by the sender.
557 * Returns 0 on success, errno otherwise
559 static int tipc_release(struct socket *sock)
561 struct sock *sk = sock->sk;
562 struct tipc_sock *tsk;
565 * Exit if socket isn't fully initialized (occurs when a failed accept()
566 * releases a pre-allocated child socket that was never used)
574 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
575 sk->sk_shutdown = SHUTDOWN_MASK;
577 tipc_sk_withdraw(tsk, 0, NULL);
578 sk_stop_timer(sk, &sk->sk_timer);
582 /* Reject any messages that accumulated in backlog queue */
584 tipc_dest_list_purge(&tsk->cong_links);
585 tsk->cong_link_cnt = 0;
586 call_rcu(&tsk->rcu, tipc_sk_callback);
593 * tipc_bind - associate or disassocate TIPC name(s) with a socket
594 * @sock: socket structure
595 * @uaddr: socket address describing name(s) and desired operation
596 * @uaddr_len: size of socket address data structure
598 * Name and name sequence binding is indicated using a positive scope value;
599 * a negative scope value unbinds the specified name. Specifying no name
600 * (i.e. a socket address length of 0) unbinds all names from the socket.
602 * Returns 0 on success, errno otherwise
604 * NOTE: This routine doesn't need to take the socket lock since it doesn't
605 * access any non-constant socket information.
607 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
610 struct sock *sk = sock->sk;
611 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
612 struct tipc_sock *tsk = tipc_sk(sk);
616 if (unlikely(!uaddr_len)) {
617 res = tipc_sk_withdraw(tsk, 0, NULL);
624 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
628 if (addr->family != AF_TIPC) {
633 if (addr->addrtype == TIPC_ADDR_NAME)
634 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
635 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
640 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
641 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
642 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
647 res = (addr->scope >= 0) ?
648 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
649 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
656 * tipc_getname - get port ID of socket or peer socket
657 * @sock: socket structure
658 * @uaddr: area for returned socket address
659 * @uaddr_len: area for returned length of socket address
660 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
662 * Returns 0 on success, errno otherwise
664 * NOTE: This routine doesn't need to take the socket lock since it only
665 * accesses socket information that is unchanging (or which changes in
666 * a completely predictable manner).
668 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
671 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
672 struct sock *sk = sock->sk;
673 struct tipc_sock *tsk = tipc_sk(sk);
675 memset(addr, 0, sizeof(*addr));
677 if ((!tipc_sk_connected(sk)) &&
678 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
680 addr->addr.id.ref = tsk_peer_port(tsk);
681 addr->addr.id.node = tsk_peer_node(tsk);
683 addr->addr.id.ref = tsk->portid;
684 addr->addr.id.node = tipc_own_addr(sock_net(sk));
687 addr->addrtype = TIPC_ADDR_ID;
688 addr->family = AF_TIPC;
690 addr->addr.name.domain = 0;
692 return sizeof(*addr);
696 * tipc_poll - read and possibly block on pollmask
697 * @file: file structure associated with the socket
698 * @sock: socket for which to calculate the poll bits
701 * Returns pollmask value
704 * It appears that the usual socket locking mechanisms are not useful here
705 * since the pollmask info is potentially out-of-date the moment this routine
706 * exits. TCP and other protocols seem to rely on higher level poll routines
707 * to handle any preventable race conditions, so TIPC will do the same ...
709 * IMPORTANT: The fact that a read or write operation is indicated does NOT
710 * imply that the operation will succeed, merely that it should be performed
711 * and will not block.
713 static __poll_t tipc_poll(struct file *file, struct socket *sock,
716 struct sock *sk = sock->sk;
717 struct tipc_sock *tsk = tipc_sk(sk);
718 __poll_t revents = 0;
720 sock_poll_wait(file, sock, wait);
722 if (sk->sk_shutdown & RCV_SHUTDOWN)
723 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
724 if (sk->sk_shutdown == SHUTDOWN_MASK)
727 switch (sk->sk_state) {
728 case TIPC_ESTABLISHED:
729 case TIPC_CONNECTING:
730 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
734 if (!skb_queue_empty(&sk->sk_receive_queue))
735 revents |= EPOLLIN | EPOLLRDNORM;
738 if (tsk->group_is_open && !tsk->cong_link_cnt)
740 if (!tipc_sk_type_connectionless(sk))
742 if (skb_queue_empty(&sk->sk_receive_queue))
744 revents |= EPOLLIN | EPOLLRDNORM;
746 case TIPC_DISCONNECTING:
747 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
754 * tipc_sendmcast - send multicast message
755 * @sock: socket structure
756 * @seq: destination address
757 * @msg: message to send
758 * @dlen: length of data to send
759 * @timeout: timeout to wait for wakeup
761 * Called from function tipc_sendmsg(), which has done all sanity checks
762 * Returns the number of bytes sent on success, or errno
764 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
765 struct msghdr *msg, size_t dlen, long timeout)
767 struct sock *sk = sock->sk;
768 struct tipc_sock *tsk = tipc_sk(sk);
769 struct tipc_msg *hdr = &tsk->phdr;
770 struct net *net = sock_net(sk);
771 int mtu = tipc_bcast_get_mtu(net);
772 struct tipc_mc_method *method = &tsk->mc_method;
773 struct sk_buff_head pkts;
774 struct tipc_nlist dsts;
780 /* Block or return if any destination link is congested */
781 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
785 /* Lookup destination nodes */
786 tipc_nlist_init(&dsts, tipc_own_addr(net));
787 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
789 if (!dsts.local && !dsts.remote)
790 return -EHOSTUNREACH;
792 /* Build message header */
793 msg_set_type(hdr, TIPC_MCAST_MSG);
794 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
795 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
796 msg_set_destport(hdr, 0);
797 msg_set_destnode(hdr, 0);
798 msg_set_nametype(hdr, seq->type);
799 msg_set_namelower(hdr, seq->lower);
800 msg_set_nameupper(hdr, seq->upper);
802 /* Build message as chain of buffers */
803 skb_queue_head_init(&pkts);
804 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
806 /* Send message if build was successful */
807 if (unlikely(rc == dlen))
808 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
809 &tsk->cong_link_cnt);
811 tipc_nlist_purge(&dsts);
813 return rc ? rc : dlen;
817 * tipc_send_group_msg - send a message to a member in the group
818 * @net: network namespace
819 * @m: message to send
821 * @dnode: destination node
822 * @dport: destination port
823 * @dlen: total length of message data
825 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
826 struct msghdr *m, struct tipc_member *mb,
827 u32 dnode, u32 dport, int dlen)
829 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
830 struct tipc_mc_method *method = &tsk->mc_method;
831 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
832 struct tipc_msg *hdr = &tsk->phdr;
833 struct sk_buff_head pkts;
836 /* Complete message header */
837 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
838 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
839 msg_set_destport(hdr, dport);
840 msg_set_destnode(hdr, dnode);
841 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
843 /* Build message as chain of buffers */
844 skb_queue_head_init(&pkts);
845 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
846 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
847 if (unlikely(rc != dlen))
851 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
852 if (unlikely(rc == -ELINKCONG)) {
853 tipc_dest_push(&tsk->cong_links, dnode, 0);
854 tsk->cong_link_cnt++;
857 /* Update send window */
858 tipc_group_update_member(mb, blks);
860 /* A broadcast sent within next EXPIRE period must follow same path */
861 method->rcast = true;
862 method->mandatory = true;
867 * tipc_send_group_unicast - send message to a member in the group
868 * @sock: socket structure
869 * @m: message to send
870 * @dlen: total length of message data
871 * @timeout: timeout to wait for wakeup
873 * Called from function tipc_sendmsg(), which has done all sanity checks
874 * Returns the number of bytes sent on success, or errno
876 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
877 int dlen, long timeout)
879 struct sock *sk = sock->sk;
880 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
881 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
882 struct tipc_sock *tsk = tipc_sk(sk);
883 struct net *net = sock_net(sk);
884 struct tipc_member *mb = NULL;
888 node = dest->addr.id.node;
889 port = dest->addr.id.ref;
891 return -EHOSTUNREACH;
893 /* Block or return if destination link or member is congested */
894 rc = tipc_wait_for_cond(sock, &timeout,
895 !tipc_dest_find(&tsk->cong_links, node, 0) &&
897 !tipc_group_cong(tsk->group, node, port, blks,
903 return -EHOSTUNREACH;
905 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
907 return rc ? rc : dlen;
911 * tipc_send_group_anycast - send message to any member with given identity
912 * @sock: socket structure
913 * @m: message to send
914 * @dlen: total length of message data
915 * @timeout: timeout to wait for wakeup
917 * Called from function tipc_sendmsg(), which has done all sanity checks
918 * Returns the number of bytes sent on success, or errno
920 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
921 int dlen, long timeout)
923 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
924 struct sock *sk = sock->sk;
925 struct tipc_sock *tsk = tipc_sk(sk);
926 struct list_head *cong_links = &tsk->cong_links;
927 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
928 struct tipc_msg *hdr = &tsk->phdr;
929 struct tipc_member *first = NULL;
930 struct tipc_member *mbr = NULL;
931 struct net *net = sock_net(sk);
932 u32 node, port, exclude;
933 struct list_head dsts;
934 u32 type, inst, scope;
939 INIT_LIST_HEAD(&dsts);
941 type = msg_nametype(hdr);
942 inst = dest->addr.name.name.instance;
943 scope = msg_lookup_scope(hdr);
945 while (++lookups < 4) {
946 exclude = tipc_group_exclude(tsk->group);
950 /* Look for a non-congested destination member, if any */
952 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
953 &dstcnt, exclude, false))
954 return -EHOSTUNREACH;
955 tipc_dest_pop(&dsts, &node, &port);
956 cong = tipc_group_cong(tsk->group, node, port, blks,
966 /* Start over if destination was not in member list */
970 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
973 /* Block or return if destination link or member is congested */
974 rc = tipc_wait_for_cond(sock, &timeout,
975 !tipc_dest_find(cong_links, node, 0) &&
977 !tipc_group_cong(tsk->group, node, port,
982 /* Send, unless destination disappeared while waiting */
987 if (unlikely(lookups >= 4))
988 return -EHOSTUNREACH;
990 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
992 return rc ? rc : dlen;
996 * tipc_send_group_bcast - send message to all members in communication group
997 * @sk: socket structure
998 * @m: message to send
999 * @dlen: total length of message data
1000 * @timeout: timeout to wait for wakeup
1002 * Called from function tipc_sendmsg(), which has done all sanity checks
1003 * Returns the number of bytes sent on success, or errno
1005 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1006 int dlen, long timeout)
1008 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1009 struct sock *sk = sock->sk;
1010 struct net *net = sock_net(sk);
1011 struct tipc_sock *tsk = tipc_sk(sk);
1012 struct tipc_nlist *dsts = tipc_group_dests(tsk->group);
1013 struct tipc_mc_method *method = &tsk->mc_method;
1014 bool ack = method->mandatory && method->rcast;
1015 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1016 struct tipc_msg *hdr = &tsk->phdr;
1017 int mtu = tipc_bcast_get_mtu(net);
1018 struct sk_buff_head pkts;
1019 int rc = -EHOSTUNREACH;
1021 if (!dsts->local && !dsts->remote)
1022 return -EHOSTUNREACH;
1024 /* Block or return if any destination link or member is congested */
1025 rc = tipc_wait_for_cond(sock, &timeout,
1026 !tsk->cong_link_cnt && tsk->group &&
1027 !tipc_group_bc_cong(tsk->group, blks));
1031 /* Complete message header */
1033 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1034 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1036 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1037 msg_set_nameinst(hdr, 0);
1039 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1040 msg_set_destport(hdr, 0);
1041 msg_set_destnode(hdr, 0);
1042 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1044 /* Avoid getting stuck with repeated forced replicasts */
1045 msg_set_grp_bc_ack_req(hdr, ack);
1047 /* Build message as chain of buffers */
1048 skb_queue_head_init(&pkts);
1049 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1050 if (unlikely(rc != dlen))
1054 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1058 /* Update broadcast sequence number and send windows */
1059 tipc_group_update_bc_members(tsk->group, blks, ack);
1061 /* Broadcast link is now free to choose method for next broadcast */
1062 method->mandatory = false;
1063 method->expires = jiffies;
1069 * tipc_send_group_mcast - send message to all members with given identity
1070 * @sock: socket structure
1071 * @m: message to send
1072 * @dlen: total length of message data
1073 * @timeout: timeout to wait for wakeup
1075 * Called from function tipc_sendmsg(), which has done all sanity checks
1076 * Returns the number of bytes sent on success, or errno
1078 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1079 int dlen, long timeout)
1081 struct sock *sk = sock->sk;
1082 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1083 struct tipc_sock *tsk = tipc_sk(sk);
1084 struct tipc_group *grp = tsk->group;
1085 struct tipc_msg *hdr = &tsk->phdr;
1086 struct net *net = sock_net(sk);
1087 u32 type, inst, scope, exclude;
1088 struct list_head dsts;
1091 INIT_LIST_HEAD(&dsts);
1093 type = msg_nametype(hdr);
1094 inst = dest->addr.name.name.instance;
1095 scope = msg_lookup_scope(hdr);
1096 exclude = tipc_group_exclude(grp);
1098 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1099 &dstcnt, exclude, true))
1100 return -EHOSTUNREACH;
1103 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1104 return tipc_send_group_unicast(sock, m, dlen, timeout);
1107 tipc_dest_list_purge(&dsts);
1108 return tipc_send_group_bcast(sock, m, dlen, timeout);
1112 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1113 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1114 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1116 * Multi-threaded: parallel calls with reference to same queues may occur
1118 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1119 struct sk_buff_head *inputq)
1121 u32 self = tipc_own_addr(net);
1122 u32 type, lower, upper, scope;
1123 struct sk_buff *skb, *_skb;
1125 struct sk_buff_head tmpq;
1126 struct list_head dports;
1127 struct tipc_msg *hdr;
1128 int user, mtyp, hlen;
1131 __skb_queue_head_init(&tmpq);
1132 INIT_LIST_HEAD(&dports);
1134 skb = tipc_skb_peek(arrvq, &inputq->lock);
1135 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1137 user = msg_user(hdr);
1138 mtyp = msg_type(hdr);
1139 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1140 onode = msg_orignode(hdr);
1141 type = msg_nametype(hdr);
1143 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1144 spin_lock_bh(&inputq->lock);
1145 if (skb_peek(arrvq) == skb) {
1146 __skb_dequeue(arrvq);
1147 __skb_queue_tail(inputq, skb);
1150 spin_unlock_bh(&inputq->lock);
1154 /* Group messages require exact scope match */
1155 if (msg_in_group(hdr)) {
1158 scope = msg_lookup_scope(hdr);
1161 /* TIPC_NODE_SCOPE means "any scope" in this context */
1163 scope = TIPC_NODE_SCOPE;
1165 scope = TIPC_CLUSTER_SCOPE;
1167 lower = msg_namelower(hdr);
1168 upper = msg_nameupper(hdr);
1171 /* Create destination port list: */
1172 tipc_nametbl_mc_lookup(net, type, lower, upper,
1173 scope, exact, &dports);
1175 /* Clone message per destination */
1176 while (tipc_dest_pop(&dports, NULL, &portid)) {
1177 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1179 msg_set_destport(buf_msg(_skb), portid);
1180 __skb_queue_tail(&tmpq, _skb);
1183 pr_warn("Failed to clone mcast rcv buffer\n");
1185 /* Append to inputq if not already done by other thread */
1186 spin_lock_bh(&inputq->lock);
1187 if (skb_peek(arrvq) == skb) {
1188 skb_queue_splice_tail_init(&tmpq, inputq);
1189 kfree_skb(__skb_dequeue(arrvq));
1191 spin_unlock_bh(&inputq->lock);
1192 __skb_queue_purge(&tmpq);
1195 tipc_sk_rcv(net, inputq);
1199 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1200 * @tsk: receiving socket
1201 * @skb: pointer to message buffer.
1203 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1204 struct sk_buff_head *inputq,
1205 struct sk_buff_head *xmitq)
1207 struct tipc_msg *hdr = buf_msg(skb);
1208 u32 onode = tsk_own_node(tsk);
1209 struct sock *sk = &tsk->sk;
1210 int mtyp = msg_type(hdr);
1213 /* Ignore if connection cannot be validated: */
1214 if (!tsk_peer_msg(tsk, hdr))
1217 if (unlikely(msg_errcode(hdr))) {
1218 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1219 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1220 tsk_peer_port(tsk));
1221 sk->sk_state_change(sk);
1223 /* State change is ignored if socket already awake,
1224 * - convert msg to abort msg and add to inqueue
1226 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1227 msg_set_type(hdr, TIPC_CONN_MSG);
1228 msg_set_size(hdr, BASIC_H_SIZE);
1229 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1230 __skb_queue_tail(inputq, skb);
1234 tsk->probe_unacked = false;
1236 if (mtyp == CONN_PROBE) {
1237 msg_set_type(hdr, CONN_PROBE_REPLY);
1238 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1239 __skb_queue_tail(xmitq, skb);
1241 } else if (mtyp == CONN_ACK) {
1242 conn_cong = tsk_conn_cong(tsk);
1243 tsk->snt_unacked -= msg_conn_ack(hdr);
1244 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1245 tsk->snd_win = msg_adv_win(hdr);
1247 sk->sk_write_space(sk);
1248 } else if (mtyp != CONN_PROBE_REPLY) {
1249 pr_warn("Received unknown CONN_PROTO msg\n");
1256 * tipc_sendmsg - send message in connectionless manner
1257 * @sock: socket structure
1258 * @m: message to send
1259 * @dsz: amount of user data to be sent
1261 * Message must have an destination specified explicitly.
1262 * Used for SOCK_RDM and SOCK_DGRAM messages,
1263 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1264 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1266 * Returns the number of bytes sent on success, or errno otherwise
1268 static int tipc_sendmsg(struct socket *sock,
1269 struct msghdr *m, size_t dsz)
1271 struct sock *sk = sock->sk;
1275 ret = __tipc_sendmsg(sock, m, dsz);
1281 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1283 struct sock *sk = sock->sk;
1284 struct net *net = sock_net(sk);
1285 struct tipc_sock *tsk = tipc_sk(sk);
1286 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1287 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1288 struct list_head *clinks = &tsk->cong_links;
1289 bool syn = !tipc_sk_type_connectionless(sk);
1290 struct tipc_group *grp = tsk->group;
1291 struct tipc_msg *hdr = &tsk->phdr;
1292 struct tipc_name_seq *seq;
1293 struct sk_buff_head pkts;
1294 u32 dport, dnode = 0;
1298 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1302 if (unlikely(m->msg_namelen < sizeof(*dest)))
1304 if (unlikely(dest->family != AF_TIPC))
1310 return tipc_send_group_bcast(sock, m, dlen, timeout);
1311 if (dest->addrtype == TIPC_ADDR_NAME)
1312 return tipc_send_group_anycast(sock, m, dlen, timeout);
1313 if (dest->addrtype == TIPC_ADDR_ID)
1314 return tipc_send_group_unicast(sock, m, dlen, timeout);
1315 if (dest->addrtype == TIPC_ADDR_MCAST)
1316 return tipc_send_group_mcast(sock, m, dlen, timeout);
1320 if (unlikely(!dest)) {
1322 if (!syn || dest->family != AF_TIPC)
1323 return -EDESTADDRREQ;
1326 if (unlikely(syn)) {
1327 if (sk->sk_state == TIPC_LISTEN)
1329 if (sk->sk_state != TIPC_OPEN)
1333 if (dest->addrtype == TIPC_ADDR_NAME) {
1334 tsk->conn_type = dest->addr.name.name.type;
1335 tsk->conn_instance = dest->addr.name.name.instance;
1337 msg_set_syn(hdr, 1);
1340 seq = &dest->addr.nameseq;
1341 if (dest->addrtype == TIPC_ADDR_MCAST)
1342 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1344 if (dest->addrtype == TIPC_ADDR_NAME) {
1345 type = dest->addr.name.name.type;
1346 inst = dest->addr.name.name.instance;
1347 dnode = dest->addr.name.domain;
1348 msg_set_type(hdr, TIPC_NAMED_MSG);
1349 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1350 msg_set_nametype(hdr, type);
1351 msg_set_nameinst(hdr, inst);
1352 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1353 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1354 msg_set_destnode(hdr, dnode);
1355 msg_set_destport(hdr, dport);
1356 if (unlikely(!dport && !dnode))
1357 return -EHOSTUNREACH;
1358 } else if (dest->addrtype == TIPC_ADDR_ID) {
1359 dnode = dest->addr.id.node;
1360 msg_set_type(hdr, TIPC_DIRECT_MSG);
1361 msg_set_lookup_scope(hdr, 0);
1362 msg_set_destnode(hdr, dnode);
1363 msg_set_destport(hdr, dest->addr.id.ref);
1364 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1369 /* Block or return if destination link is congested */
1370 rc = tipc_wait_for_cond(sock, &timeout,
1371 !tipc_dest_find(clinks, dnode, 0));
1375 skb_queue_head_init(&pkts);
1376 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1377 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1378 if (unlikely(rc != dlen))
1380 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1383 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1384 if (unlikely(rc == -ELINKCONG)) {
1385 tipc_dest_push(clinks, dnode, 0);
1386 tsk->cong_link_cnt++;
1390 if (unlikely(syn && !rc))
1391 tipc_set_sk_state(sk, TIPC_CONNECTING);
1393 return rc ? rc : dlen;
1397 * tipc_sendstream - send stream-oriented data
1398 * @sock: socket structure
1400 * @dsz: total length of data to be transmitted
1402 * Used for SOCK_STREAM data.
1404 * Returns the number of bytes sent on success (or partial success),
1405 * or errno if no data sent
1407 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1409 struct sock *sk = sock->sk;
1413 ret = __tipc_sendstream(sock, m, dsz);
1419 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1421 struct sock *sk = sock->sk;
1422 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1423 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1424 struct tipc_sock *tsk = tipc_sk(sk);
1425 struct tipc_msg *hdr = &tsk->phdr;
1426 struct net *net = sock_net(sk);
1427 struct sk_buff_head pkts;
1428 u32 dnode = tsk_peer_node(tsk);
1432 skb_queue_head_init(&pkts);
1434 if (unlikely(dlen > INT_MAX))
1437 /* Handle implicit connection setup */
1438 if (unlikely(dest)) {
1439 rc = __tipc_sendmsg(sock, m, dlen);
1440 if (dlen && dlen == rc) {
1441 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1442 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1448 rc = tipc_wait_for_cond(sock, &timeout,
1449 (!tsk->cong_link_cnt &&
1450 !tsk_conn_cong(tsk) &&
1451 tipc_sk_connected(sk)));
1455 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1456 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1457 if (unlikely(rc != send))
1460 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1461 if (unlikely(rc == -ELINKCONG)) {
1462 tsk->cong_link_cnt = 1;
1466 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1469 } while (sent < dlen && !rc);
1471 return sent ? sent : rc;
1475 * tipc_send_packet - send a connection-oriented message
1476 * @sock: socket structure
1477 * @m: message to send
1478 * @dsz: length of data to be transmitted
1480 * Used for SOCK_SEQPACKET messages.
1482 * Returns the number of bytes sent on success, or errno otherwise
1484 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1486 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1489 return tipc_sendstream(sock, m, dsz);
1492 /* tipc_sk_finish_conn - complete the setup of a connection
1494 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1497 struct sock *sk = &tsk->sk;
1498 struct net *net = sock_net(sk);
1499 struct tipc_msg *msg = &tsk->phdr;
1501 msg_set_syn(msg, 0);
1502 msg_set_destnode(msg, peer_node);
1503 msg_set_destport(msg, peer_port);
1504 msg_set_type(msg, TIPC_CONN_MSG);
1505 msg_set_lookup_scope(msg, 0);
1506 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1508 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1509 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1510 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1511 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1512 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1513 __skb_queue_purge(&sk->sk_write_queue);
1514 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1517 /* Fall back to message based flow control */
1518 tsk->rcv_win = FLOWCTL_MSG_WIN;
1519 tsk->snd_win = FLOWCTL_MSG_WIN;
1523 * tipc_sk_set_orig_addr - capture sender's address for received message
1524 * @m: descriptor for message info
1525 * @hdr: received message header
1527 * Note: Address is not captured if not requested by receiver.
1529 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1531 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1532 struct tipc_msg *hdr = buf_msg(skb);
1537 srcaddr->sock.family = AF_TIPC;
1538 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1539 srcaddr->sock.scope = 0;
1540 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1541 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1542 srcaddr->sock.addr.name.domain = 0;
1543 m->msg_namelen = sizeof(struct sockaddr_tipc);
1545 if (!msg_in_group(hdr))
1548 /* Group message users may also want to know sending member's id */
1549 srcaddr->member.family = AF_TIPC;
1550 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1551 srcaddr->member.scope = 0;
1552 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1553 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1554 srcaddr->member.addr.name.domain = 0;
1555 m->msg_namelen = sizeof(*srcaddr);
1559 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1560 * @m: descriptor for message info
1561 * @skb: received message buffer
1562 * @tsk: TIPC port associated with message
1564 * Note: Ancillary data is not captured if not requested by receiver.
1566 * Returns 0 if successful, otherwise errno
1568 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1569 struct tipc_sock *tsk)
1571 struct tipc_msg *msg;
1578 if (likely(m->msg_controllen == 0))
1582 /* Optionally capture errored message object(s) */
1583 err = msg ? msg_errcode(msg) : 0;
1584 if (unlikely(err)) {
1586 anc_data[1] = msg_data_sz(msg);
1587 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1591 if (skb_linearize(skb))
1594 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1601 /* Optionally capture message destination object */
1602 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1603 switch (dest_type) {
1604 case TIPC_NAMED_MSG:
1606 anc_data[0] = msg_nametype(msg);
1607 anc_data[1] = msg_namelower(msg);
1608 anc_data[2] = msg_namelower(msg);
1610 case TIPC_MCAST_MSG:
1612 anc_data[0] = msg_nametype(msg);
1613 anc_data[1] = msg_namelower(msg);
1614 anc_data[2] = msg_nameupper(msg);
1617 has_name = (tsk->conn_type != 0);
1618 anc_data[0] = tsk->conn_type;
1619 anc_data[1] = tsk->conn_instance;
1620 anc_data[2] = tsk->conn_instance;
1626 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1634 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1636 struct sock *sk = &tsk->sk;
1637 struct net *net = sock_net(sk);
1638 struct sk_buff *skb = NULL;
1639 struct tipc_msg *msg;
1640 u32 peer_port = tsk_peer_port(tsk);
1641 u32 dnode = tsk_peer_node(tsk);
1643 if (!tipc_sk_connected(sk))
1645 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1646 dnode, tsk_own_node(tsk), peer_port,
1647 tsk->portid, TIPC_OK);
1651 msg_set_conn_ack(msg, tsk->rcv_unacked);
1652 tsk->rcv_unacked = 0;
1654 /* Adjust to and advertize the correct window limit */
1655 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1656 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1657 msg_set_adv_win(msg, tsk->rcv_win);
1659 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1662 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1664 struct sock *sk = sock->sk;
1666 long timeo = *timeop;
1667 int err = sock_error(sk);
1673 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1674 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1675 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1680 timeo = schedule_timeout(timeo);
1684 if (!skb_queue_empty(&sk->sk_receive_queue))
1689 err = sock_intr_errno(timeo);
1690 if (signal_pending(current))
1693 err = sock_error(sk);
1697 finish_wait(sk_sleep(sk), &wait);
1703 * tipc_recvmsg - receive packet-oriented message
1704 * @m: descriptor for message info
1705 * @buflen: length of user buffer area
1706 * @flags: receive flags
1708 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1709 * If the complete message doesn't fit in user area, truncate it.
1711 * Returns size of returned message data, errno otherwise
1713 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1714 size_t buflen, int flags)
1716 struct sock *sk = sock->sk;
1717 bool connected = !tipc_sk_type_connectionless(sk);
1718 struct tipc_sock *tsk = tipc_sk(sk);
1719 int rc, err, hlen, dlen, copy;
1720 struct sk_buff_head xmitq;
1721 struct tipc_msg *hdr;
1722 struct sk_buff *skb;
1726 /* Catch invalid receive requests */
1727 if (unlikely(!buflen))
1731 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1735 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1737 /* Step rcv queue to first msg with data or error; wait if necessary */
1739 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1742 skb = skb_peek(&sk->sk_receive_queue);
1744 dlen = msg_data_sz(hdr);
1745 hlen = msg_hdr_sz(hdr);
1746 err = msg_errcode(hdr);
1747 grp_evt = msg_is_grp_evt(hdr);
1748 if (likely(dlen || err))
1750 tsk_advance_rx_queue(sk);
1753 /* Collect msg meta data, including error code and rejected data */
1754 tipc_sk_set_orig_addr(m, skb);
1755 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1760 /* Capture data if non-error msg, otherwise just set return value */
1762 copy = min_t(int, dlen, buflen);
1763 if (unlikely(copy != dlen))
1764 m->msg_flags |= MSG_TRUNC;
1765 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1769 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1775 /* Mark message as group event if applicable */
1776 if (unlikely(grp_evt)) {
1777 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1778 m->msg_flags |= MSG_EOR;
1779 m->msg_flags |= MSG_OOB;
1783 /* Caption of data or error code/rejected data was successful */
1784 if (unlikely(flags & MSG_PEEK))
1787 /* Send group flow control advertisement when applicable */
1788 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1789 skb_queue_head_init(&xmitq);
1790 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1791 msg_orignode(hdr), msg_origport(hdr),
1793 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1796 tsk_advance_rx_queue(sk);
1798 if (likely(!connected))
1801 /* Send connection flow control advertisement when applicable */
1802 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1803 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1804 tipc_sk_send_ack(tsk);
1807 return rc ? rc : copy;
1811 * tipc_recvstream - receive stream-oriented data
1812 * @m: descriptor for message info
1813 * @buflen: total size of user buffer area
1814 * @flags: receive flags
1816 * Used for SOCK_STREAM messages only. If not enough data is available
1817 * will optionally wait for more; never truncates data.
1819 * Returns size of returned message data, errno otherwise
1821 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1822 size_t buflen, int flags)
1824 struct sock *sk = sock->sk;
1825 struct tipc_sock *tsk = tipc_sk(sk);
1826 struct sk_buff *skb;
1827 struct tipc_msg *hdr;
1828 struct tipc_skb_cb *skb_cb;
1829 bool peek = flags & MSG_PEEK;
1830 int offset, required, copy, copied = 0;
1831 int hlen, dlen, err, rc;
1834 /* Catch invalid receive attempts */
1835 if (unlikely(!buflen))
1840 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1844 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1845 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1848 /* Look at first msg in receive queue; wait if necessary */
1849 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1852 skb = skb_peek(&sk->sk_receive_queue);
1853 skb_cb = TIPC_SKB_CB(skb);
1855 dlen = msg_data_sz(hdr);
1856 hlen = msg_hdr_sz(hdr);
1857 err = msg_errcode(hdr);
1859 /* Discard any empty non-errored (SYN-) message */
1860 if (unlikely(!dlen && !err)) {
1861 tsk_advance_rx_queue(sk);
1865 /* Collect msg meta data, incl. error code and rejected data */
1867 tipc_sk_set_orig_addr(m, skb);
1868 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1874 /* Copy data if msg ok, otherwise return error/partial data */
1876 offset = skb_cb->bytes_read;
1877 copy = min_t(int, dlen - offset, buflen - copied);
1878 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1883 if (unlikely(offset < dlen)) {
1885 skb_cb->bytes_read = offset;
1890 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1899 tsk_advance_rx_queue(sk);
1901 /* Send connection flow control advertisement when applicable */
1902 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1903 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1904 tipc_sk_send_ack(tsk);
1906 /* Exit if all requested data or FIN/error received */
1907 if (copied == buflen || err)
1910 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1913 return copied ? copied : rc;
1917 * tipc_write_space - wake up thread if port congestion is released
1920 static void tipc_write_space(struct sock *sk)
1922 struct socket_wq *wq;
1925 wq = rcu_dereference(sk->sk_wq);
1926 if (skwq_has_sleeper(wq))
1927 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1928 EPOLLWRNORM | EPOLLWRBAND);
1933 * tipc_data_ready - wake up threads to indicate messages have been received
1935 * @len: the length of messages
1937 static void tipc_data_ready(struct sock *sk)
1939 struct socket_wq *wq;
1942 wq = rcu_dereference(sk->sk_wq);
1943 if (skwq_has_sleeper(wq))
1944 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1945 EPOLLRDNORM | EPOLLRDBAND);
1949 static void tipc_sock_destruct(struct sock *sk)
1951 __skb_queue_purge(&sk->sk_receive_queue);
1954 static void tipc_sk_proto_rcv(struct sock *sk,
1955 struct sk_buff_head *inputq,
1956 struct sk_buff_head *xmitq)
1958 struct sk_buff *skb = __skb_dequeue(inputq);
1959 struct tipc_sock *tsk = tipc_sk(sk);
1960 struct tipc_msg *hdr = buf_msg(skb);
1961 struct tipc_group *grp = tsk->group;
1962 bool wakeup = false;
1964 switch (msg_user(hdr)) {
1966 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1969 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1970 tsk->cong_link_cnt--;
1973 case GROUP_PROTOCOL:
1974 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1977 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1978 hdr, inputq, xmitq);
1985 sk->sk_write_space(sk);
1991 * tipc_sk_filter_connect - check incoming message for a connection-based socket
1993 * @skb: pointer to message buffer.
1994 * Returns true if message should be added to receive queue, false otherwise
1996 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1998 struct sock *sk = &tsk->sk;
1999 struct net *net = sock_net(sk);
2000 struct tipc_msg *hdr = buf_msg(skb);
2001 bool con_msg = msg_connected(hdr);
2002 u32 pport = tsk_peer_port(tsk);
2003 u32 pnode = tsk_peer_node(tsk);
2004 u32 oport = msg_origport(hdr);
2005 u32 onode = msg_orignode(hdr);
2006 int err = msg_errcode(hdr);
2007 unsigned long delay;
2009 if (unlikely(msg_mcast(hdr)))
2012 switch (sk->sk_state) {
2013 case TIPC_CONNECTING:
2015 if (likely(con_msg)) {
2018 tipc_sk_finish_conn(tsk, oport, onode);
2019 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2020 /* ACK+ message with data is added to receive queue */
2021 if (msg_data_sz(hdr))
2023 /* Empty ACK-, - wake up sleeping connect() and drop */
2024 sk->sk_data_ready(sk);
2025 msg_set_dest_droppable(hdr, 1);
2028 /* Ignore connectionless message if not from listening socket */
2029 if (oport != pport || onode != pnode)
2033 if (err != TIPC_ERR_OVERLOAD)
2036 /* Prepare for new setup attempt if we have a SYN clone */
2037 if (skb_queue_empty(&sk->sk_write_queue))
2039 get_random_bytes(&delay, 2);
2040 delay %= (tsk->conn_timeout / 4);
2041 delay = msecs_to_jiffies(delay + 100);
2042 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2045 case TIPC_DISCONNECTING:
2048 /* Accept only SYN message */
2049 if (!msg_is_syn(hdr) &&
2050 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2052 if (!con_msg && !err)
2055 case TIPC_ESTABLISHED:
2056 /* Accept only connection-based messages sent by peer */
2057 if (likely(con_msg && !err && pport == oport && pnode == onode))
2059 if (!tsk_peer_msg(tsk, hdr))
2063 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2064 tipc_node_remove_conn(net, pnode, tsk->portid);
2065 sk->sk_state_change(sk);
2068 pr_err("Unknown sk_state %u\n", sk->sk_state);
2070 /* Abort connection setup attempt */
2071 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2072 sk->sk_err = ECONNREFUSED;
2073 sk->sk_state_change(sk);
2078 * rcvbuf_limit - get proper overload limit of socket receive queue
2082 * For connection oriented messages, irrespective of importance,
2083 * default queue limit is 2 MB.
2085 * For connectionless messages, queue limits are based on message
2086 * importance as follows:
2088 * TIPC_LOW_IMPORTANCE (2 MB)
2089 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2090 * TIPC_HIGH_IMPORTANCE (8 MB)
2091 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2093 * Returns overload limit according to corresponding message importance
2095 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2097 struct tipc_sock *tsk = tipc_sk(sk);
2098 struct tipc_msg *hdr = buf_msg(skb);
2100 if (unlikely(msg_in_group(hdr)))
2101 return sk->sk_rcvbuf;
2103 if (unlikely(!msg_connected(hdr)))
2104 return sk->sk_rcvbuf << msg_importance(hdr);
2106 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2107 return sk->sk_rcvbuf;
2109 return FLOWCTL_MSG_LIM;
2113 * tipc_sk_filter_rcv - validate incoming message
2115 * @skb: pointer to message.
2117 * Enqueues message on receive queue if acceptable; optionally handles
2118 * disconnect indication for a connected socket.
2120 * Called with socket lock already taken
2123 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2124 struct sk_buff_head *xmitq)
2126 bool sk_conn = !tipc_sk_type_connectionless(sk);
2127 struct tipc_sock *tsk = tipc_sk(sk);
2128 struct tipc_group *grp = tsk->group;
2129 struct tipc_msg *hdr = buf_msg(skb);
2130 struct net *net = sock_net(sk);
2131 struct sk_buff_head inputq;
2132 int limit, err = TIPC_OK;
2134 TIPC_SKB_CB(skb)->bytes_read = 0;
2135 __skb_queue_head_init(&inputq);
2136 __skb_queue_tail(&inputq, skb);
2138 if (unlikely(!msg_isdata(hdr)))
2139 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2142 tipc_group_filter_msg(grp, &inputq, xmitq);
2144 /* Validate and add to receive buffer if there is space */
2145 while ((skb = __skb_dequeue(&inputq))) {
2147 limit = rcvbuf_limit(sk, skb);
2148 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2149 (!sk_conn && msg_connected(hdr)) ||
2150 (!grp && msg_in_group(hdr)))
2151 err = TIPC_ERR_NO_PORT;
2152 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2153 atomic_inc(&sk->sk_drops);
2154 err = TIPC_ERR_OVERLOAD;
2157 if (unlikely(err)) {
2158 tipc_skb_reject(net, err, skb, xmitq);
2162 __skb_queue_tail(&sk->sk_receive_queue, skb);
2163 skb_set_owner_r(skb, sk);
2164 sk->sk_data_ready(sk);
2169 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2173 * Caller must hold socket lock
2175 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2177 unsigned int before = sk_rmem_alloc_get(sk);
2178 struct sk_buff_head xmitq;
2181 __skb_queue_head_init(&xmitq);
2183 tipc_sk_filter_rcv(sk, skb, &xmitq);
2184 added = sk_rmem_alloc_get(sk) - before;
2185 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2187 /* Send pending response/rejected messages, if any */
2188 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2193 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2194 * inputq and try adding them to socket or backlog queue
2195 * @inputq: list of incoming buffers with potentially different destinations
2196 * @sk: socket where the buffers should be enqueued
2197 * @dport: port number for the socket
2199 * Caller must hold socket lock
2201 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2202 u32 dport, struct sk_buff_head *xmitq)
2204 unsigned long time_limit = jiffies + 2;
2205 struct sk_buff *skb;
2210 while (skb_queue_len(inputq)) {
2211 if (unlikely(time_after_eq(jiffies, time_limit)))
2214 skb = tipc_skb_dequeue(inputq, dport);
2218 /* Add message directly to receive queue if possible */
2219 if (!sock_owned_by_user(sk)) {
2220 tipc_sk_filter_rcv(sk, skb, xmitq);
2224 /* Try backlog, compensating for double-counted bytes */
2225 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2226 if (!sk->sk_backlog.len)
2227 atomic_set(dcnt, 0);
2228 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2229 if (likely(!sk_add_backlog(sk, skb, lim)))
2232 /* Overload => reject message back to sender */
2233 onode = tipc_own_addr(sock_net(sk));
2234 atomic_inc(&sk->sk_drops);
2235 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2236 __skb_queue_tail(xmitq, skb);
2242 * tipc_sk_rcv - handle a chain of incoming buffers
2243 * @inputq: buffer list containing the buffers
2244 * Consumes all buffers in list until inputq is empty
2245 * Note: may be called in multiple threads referring to the same queue
2247 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2249 struct sk_buff_head xmitq;
2250 u32 dnode, dport = 0;
2252 struct tipc_sock *tsk;
2254 struct sk_buff *skb;
2256 __skb_queue_head_init(&xmitq);
2257 while (skb_queue_len(inputq)) {
2258 dport = tipc_skb_peek_port(inputq, dport);
2259 tsk = tipc_sk_lookup(net, dport);
2263 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2264 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2265 spin_unlock_bh(&sk->sk_lock.slock);
2267 /* Send pending response/rejected messages, if any */
2268 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2272 /* No destination socket => dequeue skb if still there */
2273 skb = tipc_skb_dequeue(inputq, dport);
2277 /* Try secondary lookup if unresolved named message */
2278 err = TIPC_ERR_NO_PORT;
2279 if (tipc_msg_lookup_dest(net, skb, &err))
2282 /* Prepare for message rejection */
2283 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2286 dnode = msg_destnode(buf_msg(skb));
2287 tipc_node_xmit_skb(net, skb, dnode, dport);
2291 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2293 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2294 struct sock *sk = sock->sk;
2298 int err = sock_error(sk);
2303 if (signal_pending(current))
2304 return sock_intr_errno(*timeo_p);
2306 add_wait_queue(sk_sleep(sk), &wait);
2307 done = sk_wait_event(sk, timeo_p,
2308 sk->sk_state != TIPC_CONNECTING, &wait);
2309 remove_wait_queue(sk_sleep(sk), &wait);
2315 * tipc_connect - establish a connection to another TIPC port
2316 * @sock: socket structure
2317 * @dest: socket address for destination port
2318 * @destlen: size of socket address data structure
2319 * @flags: file-related flags associated with socket
2321 * Returns 0 on success, errno otherwise
2323 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2324 int destlen, int flags)
2326 struct sock *sk = sock->sk;
2327 struct tipc_sock *tsk = tipc_sk(sk);
2328 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2329 struct msghdr m = {NULL,};
2330 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2334 if (destlen != sizeof(struct sockaddr_tipc))
2344 if (dst->family == AF_UNSPEC) {
2345 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2346 if (!tipc_sk_type_connectionless(sk))
2349 } else if (dst->family != AF_TIPC) {
2352 if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2357 /* DGRAM/RDM connect(), just save the destaddr */
2358 if (tipc_sk_type_connectionless(sk)) {
2359 memcpy(&tsk->peer, dest, destlen);
2363 previous = sk->sk_state;
2365 switch (sk->sk_state) {
2367 /* Send a 'SYN-' to destination */
2369 m.msg_namelen = destlen;
2371 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2372 * indicate send_msg() is never blocked.
2375 m.msg_flags = MSG_DONTWAIT;
2377 res = __tipc_sendmsg(sock, &m, 0);
2378 if ((res < 0) && (res != -EWOULDBLOCK))
2381 /* Just entered TIPC_CONNECTING state; the only
2382 * difference is that return value in non-blocking
2383 * case is EINPROGRESS, rather than EALREADY.
2387 case TIPC_CONNECTING:
2389 if (previous == TIPC_CONNECTING)
2393 timeout = msecs_to_jiffies(timeout);
2394 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2395 res = tipc_wait_for_connect(sock, &timeout);
2397 case TIPC_ESTABLISHED:
2410 * tipc_listen - allow socket to listen for incoming connections
2411 * @sock: socket structure
2414 * Returns 0 on success, errno otherwise
2416 static int tipc_listen(struct socket *sock, int len)
2418 struct sock *sk = sock->sk;
2422 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2428 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2430 struct sock *sk = sock->sk;
2434 /* True wake-one mechanism for incoming connections: only
2435 * one process gets woken up, not the 'whole herd'.
2436 * Since we do not 'race & poll' for established sockets
2437 * anymore, the common case will execute the loop only once.
2440 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2441 TASK_INTERRUPTIBLE);
2442 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2444 timeo = schedule_timeout(timeo);
2448 if (!skb_queue_empty(&sk->sk_receive_queue))
2453 err = sock_intr_errno(timeo);
2454 if (signal_pending(current))
2457 finish_wait(sk_sleep(sk), &wait);
2462 * tipc_accept - wait for connection request
2463 * @sock: listening socket
2464 * @newsock: new socket that is to be connected
2465 * @flags: file-related flags associated with socket
2467 * Returns 0 on success, errno otherwise
2469 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2472 struct sock *new_sk, *sk = sock->sk;
2473 struct sk_buff *buf;
2474 struct tipc_sock *new_tsock;
2475 struct tipc_msg *msg;
2481 if (sk->sk_state != TIPC_LISTEN) {
2485 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2486 res = tipc_wait_for_accept(sock, timeo);
2490 buf = skb_peek(&sk->sk_receive_queue);
2492 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2495 security_sk_clone(sock->sk, new_sock->sk);
2497 new_sk = new_sock->sk;
2498 new_tsock = tipc_sk(new_sk);
2501 /* we lock on new_sk; but lockdep sees the lock on sk */
2502 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2505 * Reject any stray messages received by new socket
2506 * before the socket lock was taken (very, very unlikely)
2508 tsk_rej_rx_queue(new_sk);
2510 /* Connect new socket to it's peer */
2511 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2513 tsk_set_importance(new_tsock, msg_importance(msg));
2514 if (msg_named(msg)) {
2515 new_tsock->conn_type = msg_nametype(msg);
2516 new_tsock->conn_instance = msg_nameinst(msg);
2520 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2521 * Respond to 'SYN+' by queuing it on new socket.
2523 if (!msg_data_sz(msg)) {
2524 struct msghdr m = {NULL,};
2526 tsk_advance_rx_queue(sk);
2527 __tipc_sendstream(new_sock, &m, 0);
2529 __skb_dequeue(&sk->sk_receive_queue);
2530 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2531 skb_set_owner_r(buf, new_sk);
2533 release_sock(new_sk);
2540 * tipc_shutdown - shutdown socket connection
2541 * @sock: socket structure
2542 * @how: direction to close (must be SHUT_RDWR)
2544 * Terminates connection (if necessary), then purges socket's receive queue.
2546 * Returns 0 on success, errno otherwise
2548 static int tipc_shutdown(struct socket *sock, int how)
2550 struct sock *sk = sock->sk;
2553 if (how != SHUT_RDWR)
2558 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2559 sk->sk_shutdown = SEND_SHUTDOWN;
2561 if (sk->sk_state == TIPC_DISCONNECTING) {
2562 /* Discard any unreceived messages */
2563 __skb_queue_purge(&sk->sk_receive_queue);
2565 /* Wake up anyone sleeping in poll */
2566 sk->sk_state_change(sk);
2576 static void tipc_sk_check_probing_state(struct sock *sk,
2577 struct sk_buff_head *list)
2579 struct tipc_sock *tsk = tipc_sk(sk);
2580 u32 pnode = tsk_peer_node(tsk);
2581 u32 pport = tsk_peer_port(tsk);
2582 u32 self = tsk_own_node(tsk);
2583 u32 oport = tsk->portid;
2584 struct sk_buff *skb;
2586 if (tsk->probe_unacked) {
2587 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2588 sk->sk_err = ECONNABORTED;
2589 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2590 sk->sk_state_change(sk);
2593 /* Prepare new probe */
2594 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2595 pnode, self, pport, oport, TIPC_OK);
2597 __skb_queue_tail(list, skb);
2598 tsk->probe_unacked = true;
2599 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2602 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2604 struct tipc_sock *tsk = tipc_sk(sk);
2606 /* Try again later if dest link is congested */
2607 if (tsk->cong_link_cnt) {
2608 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2611 /* Prepare SYN for retransmit */
2612 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2615 static void tipc_sk_timeout(struct timer_list *t)
2617 struct sock *sk = from_timer(sk, t, sk_timer);
2618 struct tipc_sock *tsk = tipc_sk(sk);
2619 u32 pnode = tsk_peer_node(tsk);
2620 struct sk_buff_head list;
2623 skb_queue_head_init(&list);
2626 /* Try again later if socket is busy */
2627 if (sock_owned_by_user(sk)) {
2628 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2633 if (sk->sk_state == TIPC_ESTABLISHED)
2634 tipc_sk_check_probing_state(sk, &list);
2635 else if (sk->sk_state == TIPC_CONNECTING)
2636 tipc_sk_retry_connect(sk, &list);
2640 if (!skb_queue_empty(&list))
2641 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2643 /* SYN messages may cause link congestion */
2644 if (rc == -ELINKCONG) {
2645 tipc_dest_push(&tsk->cong_links, pnode, 0);
2646 tsk->cong_link_cnt = 1;
2651 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2652 struct tipc_name_seq const *seq)
2654 struct sock *sk = &tsk->sk;
2655 struct net *net = sock_net(sk);
2656 struct publication *publ;
2659 if (scope != TIPC_NODE_SCOPE)
2660 scope = TIPC_CLUSTER_SCOPE;
2662 if (tipc_sk_connected(sk))
2664 key = tsk->portid + tsk->pub_count + 1;
2665 if (key == tsk->portid)
2668 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2669 scope, tsk->portid, key);
2670 if (unlikely(!publ))
2673 list_add(&publ->binding_sock, &tsk->publications);
2679 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2680 struct tipc_name_seq const *seq)
2682 struct net *net = sock_net(&tsk->sk);
2683 struct publication *publ;
2684 struct publication *safe;
2687 if (scope != TIPC_NODE_SCOPE)
2688 scope = TIPC_CLUSTER_SCOPE;
2690 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2692 if (publ->scope != scope)
2694 if (publ->type != seq->type)
2696 if (publ->lower != seq->lower)
2698 if (publ->upper != seq->upper)
2700 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2701 publ->upper, publ->key);
2705 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2706 publ->upper, publ->key);
2709 if (list_empty(&tsk->publications))
2714 /* tipc_sk_reinit: set non-zero address in all existing sockets
2715 * when we go from standalone to network mode.
2717 void tipc_sk_reinit(struct net *net)
2719 struct tipc_net *tn = net_generic(net, tipc_net_id);
2720 struct rhashtable_iter iter;
2721 struct tipc_sock *tsk;
2722 struct tipc_msg *msg;
2724 rhashtable_walk_enter(&tn->sk_rht, &iter);
2727 rhashtable_walk_start(&iter);
2729 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2730 sock_hold(&tsk->sk);
2731 rhashtable_walk_stop(&iter);
2732 lock_sock(&tsk->sk);
2734 msg_set_prevnode(msg, tipc_own_addr(net));
2735 msg_set_orignode(msg, tipc_own_addr(net));
2736 release_sock(&tsk->sk);
2737 rhashtable_walk_start(&iter);
2741 rhashtable_walk_stop(&iter);
2742 } while (tsk == ERR_PTR(-EAGAIN));
2744 rhashtable_walk_exit(&iter);
2747 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2749 struct tipc_net *tn = net_generic(net, tipc_net_id);
2750 struct tipc_sock *tsk;
2753 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2755 sock_hold(&tsk->sk);
2761 static int tipc_sk_insert(struct tipc_sock *tsk)
2763 struct sock *sk = &tsk->sk;
2764 struct net *net = sock_net(sk);
2765 struct tipc_net *tn = net_generic(net, tipc_net_id);
2766 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2767 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2769 while (remaining--) {
2771 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2772 portid = TIPC_MIN_PORT;
2773 tsk->portid = portid;
2774 sock_hold(&tsk->sk);
2775 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2784 static void tipc_sk_remove(struct tipc_sock *tsk)
2786 struct sock *sk = &tsk->sk;
2787 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2789 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2790 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2795 static const struct rhashtable_params tsk_rht_params = {
2797 .head_offset = offsetof(struct tipc_sock, node),
2798 .key_offset = offsetof(struct tipc_sock, portid),
2799 .key_len = sizeof(u32), /* portid */
2800 .max_size = 1048576,
2802 .automatic_shrinking = true,
2805 int tipc_sk_rht_init(struct net *net)
2807 struct tipc_net *tn = net_generic(net, tipc_net_id);
2809 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2812 void tipc_sk_rht_destroy(struct net *net)
2814 struct tipc_net *tn = net_generic(net, tipc_net_id);
2816 /* Wait for socket readers to complete */
2819 rhashtable_destroy(&tn->sk_rht);
2822 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2824 struct net *net = sock_net(&tsk->sk);
2825 struct tipc_group *grp = tsk->group;
2826 struct tipc_msg *hdr = &tsk->phdr;
2827 struct tipc_name_seq seq;
2830 if (mreq->type < TIPC_RESERVED_TYPES)
2832 if (mreq->scope > TIPC_NODE_SCOPE)
2836 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2840 msg_set_lookup_scope(hdr, mreq->scope);
2841 msg_set_nametype(hdr, mreq->type);
2842 msg_set_dest_droppable(hdr, true);
2843 seq.type = mreq->type;
2844 seq.lower = mreq->instance;
2845 seq.upper = seq.lower;
2846 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2847 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2849 tipc_group_delete(net, grp);
2853 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2854 tsk->mc_method.rcast = true;
2855 tsk->mc_method.mandatory = true;
2856 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2860 static int tipc_sk_leave(struct tipc_sock *tsk)
2862 struct net *net = sock_net(&tsk->sk);
2863 struct tipc_group *grp = tsk->group;
2864 struct tipc_name_seq seq;
2869 tipc_group_self(grp, &seq, &scope);
2870 tipc_group_delete(net, grp);
2872 tipc_sk_withdraw(tsk, scope, &seq);
2877 * tipc_setsockopt - set socket option
2878 * @sock: socket structure
2879 * @lvl: option level
2880 * @opt: option identifier
2881 * @ov: pointer to new option value
2882 * @ol: length of option value
2884 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2885 * (to ease compatibility).
2887 * Returns 0 on success, errno otherwise
2889 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2890 char __user *ov, unsigned int ol)
2892 struct sock *sk = sock->sk;
2893 struct tipc_sock *tsk = tipc_sk(sk);
2894 struct tipc_group_req mreq;
2898 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2900 if (lvl != SOL_TIPC)
2901 return -ENOPROTOOPT;
2904 case TIPC_IMPORTANCE:
2905 case TIPC_SRC_DROPPABLE:
2906 case TIPC_DEST_DROPPABLE:
2907 case TIPC_CONN_TIMEOUT:
2908 if (ol < sizeof(value))
2910 if (get_user(value, (u32 __user *)ov))
2913 case TIPC_GROUP_JOIN:
2914 if (ol < sizeof(mreq))
2916 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2927 case TIPC_IMPORTANCE:
2928 res = tsk_set_importance(tsk, value);
2930 case TIPC_SRC_DROPPABLE:
2931 if (sock->type != SOCK_STREAM)
2932 tsk_set_unreliable(tsk, value);
2936 case TIPC_DEST_DROPPABLE:
2937 tsk_set_unreturnable(tsk, value);
2939 case TIPC_CONN_TIMEOUT:
2940 tipc_sk(sk)->conn_timeout = value;
2942 case TIPC_MCAST_BROADCAST:
2943 tsk->mc_method.rcast = false;
2944 tsk->mc_method.mandatory = true;
2946 case TIPC_MCAST_REPLICAST:
2947 tsk->mc_method.rcast = true;
2948 tsk->mc_method.mandatory = true;
2950 case TIPC_GROUP_JOIN:
2951 res = tipc_sk_join(tsk, &mreq);
2953 case TIPC_GROUP_LEAVE:
2954 res = tipc_sk_leave(tsk);
2966 * tipc_getsockopt - get socket option
2967 * @sock: socket structure
2968 * @lvl: option level
2969 * @opt: option identifier
2970 * @ov: receptacle for option value
2971 * @ol: receptacle for length of option value
2973 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2974 * (to ease compatibility).
2976 * Returns 0 on success, errno otherwise
2978 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2979 char __user *ov, int __user *ol)
2981 struct sock *sk = sock->sk;
2982 struct tipc_sock *tsk = tipc_sk(sk);
2983 struct tipc_name_seq seq;
2988 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2989 return put_user(0, ol);
2990 if (lvl != SOL_TIPC)
2991 return -ENOPROTOOPT;
2992 res = get_user(len, ol);
2999 case TIPC_IMPORTANCE:
3000 value = tsk_importance(tsk);
3002 case TIPC_SRC_DROPPABLE:
3003 value = tsk_unreliable(tsk);
3005 case TIPC_DEST_DROPPABLE:
3006 value = tsk_unreturnable(tsk);
3008 case TIPC_CONN_TIMEOUT:
3009 value = tsk->conn_timeout;
3010 /* no need to set "res", since already 0 at this point */
3012 case TIPC_NODE_RECVQ_DEPTH:
3013 value = 0; /* was tipc_queue_size, now obsolete */
3015 case TIPC_SOCK_RECVQ_DEPTH:
3016 value = skb_queue_len(&sk->sk_receive_queue);
3018 case TIPC_GROUP_JOIN:
3021 tipc_group_self(tsk->group, &seq, &scope);
3031 return res; /* "get" failed */
3033 if (len < sizeof(value))
3036 if (copy_to_user(ov, &value, sizeof(value)))
3039 return put_user(sizeof(value), ol);
3042 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3044 struct net *net = sock_net(sock->sk);
3045 struct tipc_sioc_nodeid_req nr = {0};
3046 struct tipc_sioc_ln_req lnr;
3047 void __user *argp = (void __user *)arg;
3050 case SIOCGETLINKNAME:
3051 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3053 if (!tipc_node_get_linkname(net,
3054 lnr.bearer_id & 0xffff, lnr.peer,
3055 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3056 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3060 return -EADDRNOTAVAIL;
3062 if (copy_from_user(&nr, argp, sizeof(nr)))
3064 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3065 return -EADDRNOTAVAIL;
3066 if (copy_to_user(argp, &nr, sizeof(nr)))
3070 return -ENOIOCTLCMD;
3074 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3076 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3077 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3078 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3080 tsk1->peer.family = AF_TIPC;
3081 tsk1->peer.addrtype = TIPC_ADDR_ID;
3082 tsk1->peer.scope = TIPC_NODE_SCOPE;
3083 tsk1->peer.addr.id.ref = tsk2->portid;
3084 tsk1->peer.addr.id.node = onode;
3085 tsk2->peer.family = AF_TIPC;
3086 tsk2->peer.addrtype = TIPC_ADDR_ID;
3087 tsk2->peer.scope = TIPC_NODE_SCOPE;
3088 tsk2->peer.addr.id.ref = tsk1->portid;
3089 tsk2->peer.addr.id.node = onode;
3091 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3092 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3096 /* Protocol switches for the various types of TIPC sockets */
3098 static const struct proto_ops msg_ops = {
3099 .owner = THIS_MODULE,
3101 .release = tipc_release,
3103 .connect = tipc_connect,
3104 .socketpair = tipc_socketpair,
3105 .accept = sock_no_accept,
3106 .getname = tipc_getname,
3108 .ioctl = tipc_ioctl,
3109 .listen = sock_no_listen,
3110 .shutdown = tipc_shutdown,
3111 .setsockopt = tipc_setsockopt,
3112 .getsockopt = tipc_getsockopt,
3113 .sendmsg = tipc_sendmsg,
3114 .recvmsg = tipc_recvmsg,
3115 .mmap = sock_no_mmap,
3116 .sendpage = sock_no_sendpage
3119 static const struct proto_ops packet_ops = {
3120 .owner = THIS_MODULE,
3122 .release = tipc_release,
3124 .connect = tipc_connect,
3125 .socketpair = tipc_socketpair,
3126 .accept = tipc_accept,
3127 .getname = tipc_getname,
3129 .ioctl = tipc_ioctl,
3130 .listen = tipc_listen,
3131 .shutdown = tipc_shutdown,
3132 .setsockopt = tipc_setsockopt,
3133 .getsockopt = tipc_getsockopt,
3134 .sendmsg = tipc_send_packet,
3135 .recvmsg = tipc_recvmsg,
3136 .mmap = sock_no_mmap,
3137 .sendpage = sock_no_sendpage
3140 static const struct proto_ops stream_ops = {
3141 .owner = THIS_MODULE,
3143 .release = tipc_release,
3145 .connect = tipc_connect,
3146 .socketpair = tipc_socketpair,
3147 .accept = tipc_accept,
3148 .getname = tipc_getname,
3150 .ioctl = tipc_ioctl,
3151 .listen = tipc_listen,
3152 .shutdown = tipc_shutdown,
3153 .setsockopt = tipc_setsockopt,
3154 .getsockopt = tipc_getsockopt,
3155 .sendmsg = tipc_sendstream,
3156 .recvmsg = tipc_recvstream,
3157 .mmap = sock_no_mmap,
3158 .sendpage = sock_no_sendpage
3161 static const struct net_proto_family tipc_family_ops = {
3162 .owner = THIS_MODULE,
3164 .create = tipc_sk_create
3167 static struct proto tipc_proto = {
3169 .owner = THIS_MODULE,
3170 .obj_size = sizeof(struct tipc_sock),
3171 .sysctl_rmem = sysctl_tipc_rmem
3175 * tipc_socket_init - initialize TIPC socket interface
3177 * Returns 0 on success, errno otherwise
3179 int tipc_socket_init(void)
3183 res = proto_register(&tipc_proto, 1);
3185 pr_err("Failed to register TIPC protocol type\n");
3189 res = sock_register(&tipc_family_ops);
3191 pr_err("Failed to register TIPC socket type\n");
3192 proto_unregister(&tipc_proto);
3200 * tipc_socket_stop - stop TIPC socket interface
3202 void tipc_socket_stop(void)
3204 sock_unregister(tipc_family_ops.family);
3205 proto_unregister(&tipc_proto);
3208 /* Caller should hold socket lock for the passed tipc socket. */
3209 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3213 struct nlattr *nest;
3215 peer_node = tsk_peer_node(tsk);
3216 peer_port = tsk_peer_port(tsk);
3218 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3220 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3222 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3225 if (tsk->conn_type != 0) {
3226 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3228 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3230 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3233 nla_nest_end(skb, nest);
3238 nla_nest_cancel(skb, nest);
3243 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3246 struct net *net = sock_net(skb->sk);
3247 struct sock *sk = &tsk->sk;
3249 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3250 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3253 if (tipc_sk_connected(sk)) {
3254 if (__tipc_nl_add_sk_con(skb, tsk))
3256 } else if (!list_empty(&tsk->publications)) {
3257 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3263 /* Caller should hold socket lock for the passed tipc socket. */
3264 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3265 struct tipc_sock *tsk)
3267 struct nlattr *attrs;
3270 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3271 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3275 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3277 goto genlmsg_cancel;
3279 if (__tipc_nl_add_sk_info(skb, tsk))
3280 goto attr_msg_cancel;
3282 nla_nest_end(skb, attrs);
3283 genlmsg_end(skb, hdr);
3288 nla_nest_cancel(skb, attrs);
3290 genlmsg_cancel(skb, hdr);
3295 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3296 int (*skb_handler)(struct sk_buff *skb,
3297 struct netlink_callback *cb,
3298 struct tipc_sock *tsk))
3300 struct rhashtable_iter *iter = (void *)cb->args[4];
3301 struct tipc_sock *tsk;
3304 rhashtable_walk_start(iter);
3305 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3308 if (err == -EAGAIN) {
3315 sock_hold(&tsk->sk);
3316 rhashtable_walk_stop(iter);
3317 lock_sock(&tsk->sk);
3318 err = skb_handler(skb, cb, tsk);
3320 release_sock(&tsk->sk);
3324 release_sock(&tsk->sk);
3325 rhashtable_walk_start(iter);
3328 rhashtable_walk_stop(iter);
3332 EXPORT_SYMBOL(tipc_nl_sk_walk);
3334 int tipc_dump_start(struct netlink_callback *cb)
3336 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3338 EXPORT_SYMBOL(tipc_dump_start);
3340 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3342 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3343 struct rhashtable_iter *iter = (void *)cb->args[4];
3344 struct tipc_net *tn = tipc_net(net);
3347 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3351 cb->args[4] = (long)iter;
3354 rhashtable_walk_enter(&tn->sk_rht, iter);
3358 int tipc_dump_done(struct netlink_callback *cb)
3360 struct rhashtable_iter *hti = (void *)cb->args[4];
3362 rhashtable_walk_exit(hti);
3366 EXPORT_SYMBOL(tipc_dump_done);
3368 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3369 struct tipc_sock *tsk, u32 sk_filter_state,
3370 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3372 struct sock *sk = &tsk->sk;
3373 struct nlattr *attrs;
3374 struct nlattr *stat;
3376 /*filter response w.r.t sk_state*/
3377 if (!(sk_filter_state & (1 << sk->sk_state)))
3380 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3384 if (__tipc_nl_add_sk_info(skb, tsk))
3385 goto attr_msg_cancel;
3387 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3388 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3389 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3390 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3391 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3393 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3394 tipc_diag_gen_cookie(sk),
3396 goto attr_msg_cancel;
3398 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3400 goto attr_msg_cancel;
3402 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3403 skb_queue_len(&sk->sk_receive_queue)) ||
3404 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3405 skb_queue_len(&sk->sk_write_queue)) ||
3406 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3407 atomic_read(&sk->sk_drops)))
3408 goto stat_msg_cancel;
3410 if (tsk->cong_link_cnt &&
3411 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3412 goto stat_msg_cancel;
3414 if (tsk_conn_cong(tsk) &&
3415 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3416 goto stat_msg_cancel;
3418 nla_nest_end(skb, stat);
3421 if (tipc_group_fill_sock_diag(tsk->group, skb))
3422 goto stat_msg_cancel;
3424 nla_nest_end(skb, attrs);
3429 nla_nest_cancel(skb, stat);
3431 nla_nest_cancel(skb, attrs);
3435 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3437 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3439 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3442 /* Caller should hold socket lock for the passed tipc socket. */
3443 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3444 struct netlink_callback *cb,
3445 struct publication *publ)
3448 struct nlattr *attrs;
3450 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3451 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3455 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3457 goto genlmsg_cancel;
3459 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3460 goto attr_msg_cancel;
3461 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3462 goto attr_msg_cancel;
3463 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3464 goto attr_msg_cancel;
3465 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3466 goto attr_msg_cancel;
3468 nla_nest_end(skb, attrs);
3469 genlmsg_end(skb, hdr);
3474 nla_nest_cancel(skb, attrs);
3476 genlmsg_cancel(skb, hdr);
3481 /* Caller should hold socket lock for the passed tipc socket. */
3482 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3483 struct netlink_callback *cb,
3484 struct tipc_sock *tsk, u32 *last_publ)
3487 struct publication *p;
3490 list_for_each_entry(p, &tsk->publications, binding_sock) {
3491 if (p->key == *last_publ)
3494 if (p->key != *last_publ) {
3495 /* We never set seq or call nl_dump_check_consistent()
3496 * this means that setting prev_seq here will cause the
3497 * consistence check to fail in the netlink callback
3498 * handler. Resulting in the last NLMSG_DONE message
3499 * having the NLM_F_DUMP_INTR flag set.
3506 p = list_first_entry(&tsk->publications, struct publication,
3510 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3511 err = __tipc_nl_add_sk_publ(skb, cb, p);
3513 *last_publ = p->key;
3522 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3525 u32 tsk_portid = cb->args[0];
3526 u32 last_publ = cb->args[1];
3527 u32 done = cb->args[2];
3528 struct net *net = sock_net(skb->sk);
3529 struct tipc_sock *tsk;
3532 struct nlattr **attrs;
3533 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3535 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3539 if (!attrs[TIPC_NLA_SOCK])
3542 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3543 attrs[TIPC_NLA_SOCK],
3544 tipc_nl_sock_policy, NULL);
3548 if (!sock[TIPC_NLA_SOCK_REF])
3551 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3557 tsk = tipc_sk_lookup(net, tsk_portid);
3561 lock_sock(&tsk->sk);
3562 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3565 release_sock(&tsk->sk);
3568 cb->args[0] = tsk_portid;
3569 cb->args[1] = last_publ;