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 tipc_group *grp = tsk->group;
884 struct net *net = sock_net(sk);
885 struct tipc_member *mb = NULL;
889 node = dest->addr.id.node;
890 port = dest->addr.id.ref;
892 return -EHOSTUNREACH;
894 /* Block or return if destination link or member is congested */
895 rc = tipc_wait_for_cond(sock, &timeout,
896 !tipc_dest_find(&tsk->cong_links, node, 0) &&
897 !tipc_group_cong(grp, node, port, blks, &mb));
902 return -EHOSTUNREACH;
904 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
906 return rc ? rc : dlen;
910 * tipc_send_group_anycast - send message to any member with given identity
911 * @sock: socket structure
912 * @m: message to send
913 * @dlen: total length of message data
914 * @timeout: timeout to wait for wakeup
916 * Called from function tipc_sendmsg(), which has done all sanity checks
917 * Returns the number of bytes sent on success, or errno
919 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
920 int dlen, long timeout)
922 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
923 struct sock *sk = sock->sk;
924 struct tipc_sock *tsk = tipc_sk(sk);
925 struct list_head *cong_links = &tsk->cong_links;
926 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
927 struct tipc_group *grp = tsk->group;
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);
944 exclude = tipc_group_exclude(grp);
946 while (++lookups < 4) {
949 /* Look for a non-congested destination member, if any */
951 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
952 &dstcnt, exclude, false))
953 return -EHOSTUNREACH;
954 tipc_dest_pop(&dsts, &node, &port);
955 cong = tipc_group_cong(grp, node, port, blks, &mbr);
964 /* Start over if destination was not in member list */
968 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
971 /* Block or return if destination link or member is congested */
972 rc = tipc_wait_for_cond(sock, &timeout,
973 !tipc_dest_find(cong_links, node, 0) &&
974 !tipc_group_cong(grp, node, port,
979 /* Send, unless destination disappeared while waiting */
984 if (unlikely(lookups >= 4))
985 return -EHOSTUNREACH;
987 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
989 return rc ? rc : dlen;
993 * tipc_send_group_bcast - send message to all members in communication group
994 * @sk: socket structure
995 * @m: message to send
996 * @dlen: total length of message data
997 * @timeout: timeout to wait for wakeup
999 * Called from function tipc_sendmsg(), which has done all sanity checks
1000 * Returns the number of bytes sent on success, or errno
1002 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1003 int dlen, long timeout)
1005 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1006 struct sock *sk = sock->sk;
1007 struct net *net = sock_net(sk);
1008 struct tipc_sock *tsk = tipc_sk(sk);
1009 struct tipc_group *grp = tsk->group;
1010 struct tipc_nlist *dsts = tipc_group_dests(grp);
1011 struct tipc_mc_method *method = &tsk->mc_method;
1012 bool ack = method->mandatory && method->rcast;
1013 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1014 struct tipc_msg *hdr = &tsk->phdr;
1015 int mtu = tipc_bcast_get_mtu(net);
1016 struct sk_buff_head pkts;
1017 int rc = -EHOSTUNREACH;
1019 if (!dsts->local && !dsts->remote)
1020 return -EHOSTUNREACH;
1022 /* Block or return if any destination link or member is congested */
1023 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt &&
1024 !tipc_group_bc_cong(grp, blks));
1028 /* Complete message header */
1030 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1031 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1033 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1034 msg_set_nameinst(hdr, 0);
1036 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1037 msg_set_destport(hdr, 0);
1038 msg_set_destnode(hdr, 0);
1039 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1041 /* Avoid getting stuck with repeated forced replicasts */
1042 msg_set_grp_bc_ack_req(hdr, ack);
1044 /* Build message as chain of buffers */
1045 skb_queue_head_init(&pkts);
1046 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1047 if (unlikely(rc != dlen))
1051 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1055 /* Update broadcast sequence number and send windows */
1056 tipc_group_update_bc_members(tsk->group, blks, ack);
1058 /* Broadcast link is now free to choose method for next broadcast */
1059 method->mandatory = false;
1060 method->expires = jiffies;
1066 * tipc_send_group_mcast - send message to all members with given identity
1067 * @sock: socket structure
1068 * @m: message to send
1069 * @dlen: total length of message data
1070 * @timeout: timeout to wait for wakeup
1072 * Called from function tipc_sendmsg(), which has done all sanity checks
1073 * Returns the number of bytes sent on success, or errno
1075 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1076 int dlen, long timeout)
1078 struct sock *sk = sock->sk;
1079 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1080 struct tipc_sock *tsk = tipc_sk(sk);
1081 struct tipc_group *grp = tsk->group;
1082 struct tipc_msg *hdr = &tsk->phdr;
1083 struct net *net = sock_net(sk);
1084 u32 type, inst, scope, exclude;
1085 struct list_head dsts;
1088 INIT_LIST_HEAD(&dsts);
1090 type = msg_nametype(hdr);
1091 inst = dest->addr.name.name.instance;
1092 scope = msg_lookup_scope(hdr);
1093 exclude = tipc_group_exclude(grp);
1095 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1096 &dstcnt, exclude, true))
1097 return -EHOSTUNREACH;
1100 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1101 return tipc_send_group_unicast(sock, m, dlen, timeout);
1104 tipc_dest_list_purge(&dsts);
1105 return tipc_send_group_bcast(sock, m, dlen, timeout);
1109 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1110 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1111 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1113 * Multi-threaded: parallel calls with reference to same queues may occur
1115 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1116 struct sk_buff_head *inputq)
1118 u32 self = tipc_own_addr(net);
1119 u32 type, lower, upper, scope;
1120 struct sk_buff *skb, *_skb;
1122 struct sk_buff_head tmpq;
1123 struct list_head dports;
1124 struct tipc_msg *hdr;
1125 int user, mtyp, hlen;
1128 __skb_queue_head_init(&tmpq);
1129 INIT_LIST_HEAD(&dports);
1131 skb = tipc_skb_peek(arrvq, &inputq->lock);
1132 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1134 user = msg_user(hdr);
1135 mtyp = msg_type(hdr);
1136 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1137 onode = msg_orignode(hdr);
1138 type = msg_nametype(hdr);
1140 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1141 spin_lock_bh(&inputq->lock);
1142 if (skb_peek(arrvq) == skb) {
1143 __skb_dequeue(arrvq);
1144 __skb_queue_tail(inputq, skb);
1147 spin_unlock_bh(&inputq->lock);
1151 /* Group messages require exact scope match */
1152 if (msg_in_group(hdr)) {
1155 scope = msg_lookup_scope(hdr);
1158 /* TIPC_NODE_SCOPE means "any scope" in this context */
1160 scope = TIPC_NODE_SCOPE;
1162 scope = TIPC_CLUSTER_SCOPE;
1164 lower = msg_namelower(hdr);
1165 upper = msg_nameupper(hdr);
1168 /* Create destination port list: */
1169 tipc_nametbl_mc_lookup(net, type, lower, upper,
1170 scope, exact, &dports);
1172 /* Clone message per destination */
1173 while (tipc_dest_pop(&dports, NULL, &portid)) {
1174 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1176 msg_set_destport(buf_msg(_skb), portid);
1177 __skb_queue_tail(&tmpq, _skb);
1180 pr_warn("Failed to clone mcast rcv buffer\n");
1182 /* Append to inputq if not already done by other thread */
1183 spin_lock_bh(&inputq->lock);
1184 if (skb_peek(arrvq) == skb) {
1185 skb_queue_splice_tail_init(&tmpq, inputq);
1186 kfree_skb(__skb_dequeue(arrvq));
1188 spin_unlock_bh(&inputq->lock);
1189 __skb_queue_purge(&tmpq);
1192 tipc_sk_rcv(net, inputq);
1196 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1197 * @tsk: receiving socket
1198 * @skb: pointer to message buffer.
1200 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1201 struct sk_buff_head *inputq,
1202 struct sk_buff_head *xmitq)
1204 struct tipc_msg *hdr = buf_msg(skb);
1205 u32 onode = tsk_own_node(tsk);
1206 struct sock *sk = &tsk->sk;
1207 int mtyp = msg_type(hdr);
1210 /* Ignore if connection cannot be validated: */
1211 if (!tsk_peer_msg(tsk, hdr))
1214 if (unlikely(msg_errcode(hdr))) {
1215 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1216 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1217 tsk_peer_port(tsk));
1218 sk->sk_state_change(sk);
1220 /* State change is ignored if socket already awake,
1221 * - convert msg to abort msg and add to inqueue
1223 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1224 msg_set_type(hdr, TIPC_CONN_MSG);
1225 msg_set_size(hdr, BASIC_H_SIZE);
1226 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1227 __skb_queue_tail(inputq, skb);
1231 tsk->probe_unacked = false;
1233 if (mtyp == CONN_PROBE) {
1234 msg_set_type(hdr, CONN_PROBE_REPLY);
1235 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1236 __skb_queue_tail(xmitq, skb);
1238 } else if (mtyp == CONN_ACK) {
1239 conn_cong = tsk_conn_cong(tsk);
1240 tsk->snt_unacked -= msg_conn_ack(hdr);
1241 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1242 tsk->snd_win = msg_adv_win(hdr);
1244 sk->sk_write_space(sk);
1245 } else if (mtyp != CONN_PROBE_REPLY) {
1246 pr_warn("Received unknown CONN_PROTO msg\n");
1253 * tipc_sendmsg - send message in connectionless manner
1254 * @sock: socket structure
1255 * @m: message to send
1256 * @dsz: amount of user data to be sent
1258 * Message must have an destination specified explicitly.
1259 * Used for SOCK_RDM and SOCK_DGRAM messages,
1260 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1261 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1263 * Returns the number of bytes sent on success, or errno otherwise
1265 static int tipc_sendmsg(struct socket *sock,
1266 struct msghdr *m, size_t dsz)
1268 struct sock *sk = sock->sk;
1272 ret = __tipc_sendmsg(sock, m, dsz);
1278 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1280 struct sock *sk = sock->sk;
1281 struct net *net = sock_net(sk);
1282 struct tipc_sock *tsk = tipc_sk(sk);
1283 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1284 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1285 struct list_head *clinks = &tsk->cong_links;
1286 bool syn = !tipc_sk_type_connectionless(sk);
1287 struct tipc_group *grp = tsk->group;
1288 struct tipc_msg *hdr = &tsk->phdr;
1289 struct tipc_name_seq *seq;
1290 struct sk_buff_head pkts;
1291 u32 dport, dnode = 0;
1295 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1299 if (unlikely(m->msg_namelen < sizeof(*dest)))
1301 if (unlikely(dest->family != AF_TIPC))
1307 return tipc_send_group_bcast(sock, m, dlen, timeout);
1308 if (dest->addrtype == TIPC_ADDR_NAME)
1309 return tipc_send_group_anycast(sock, m, dlen, timeout);
1310 if (dest->addrtype == TIPC_ADDR_ID)
1311 return tipc_send_group_unicast(sock, m, dlen, timeout);
1312 if (dest->addrtype == TIPC_ADDR_MCAST)
1313 return tipc_send_group_mcast(sock, m, dlen, timeout);
1317 if (unlikely(!dest)) {
1319 if (!syn || dest->family != AF_TIPC)
1320 return -EDESTADDRREQ;
1323 if (unlikely(syn)) {
1324 if (sk->sk_state == TIPC_LISTEN)
1326 if (sk->sk_state != TIPC_OPEN)
1330 if (dest->addrtype == TIPC_ADDR_NAME) {
1331 tsk->conn_type = dest->addr.name.name.type;
1332 tsk->conn_instance = dest->addr.name.name.instance;
1334 msg_set_syn(hdr, 1);
1337 seq = &dest->addr.nameseq;
1338 if (dest->addrtype == TIPC_ADDR_MCAST)
1339 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1341 if (dest->addrtype == TIPC_ADDR_NAME) {
1342 type = dest->addr.name.name.type;
1343 inst = dest->addr.name.name.instance;
1344 dnode = dest->addr.name.domain;
1345 msg_set_type(hdr, TIPC_NAMED_MSG);
1346 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1347 msg_set_nametype(hdr, type);
1348 msg_set_nameinst(hdr, inst);
1349 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1350 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1351 msg_set_destnode(hdr, dnode);
1352 msg_set_destport(hdr, dport);
1353 if (unlikely(!dport && !dnode))
1354 return -EHOSTUNREACH;
1355 } else if (dest->addrtype == TIPC_ADDR_ID) {
1356 dnode = dest->addr.id.node;
1357 msg_set_type(hdr, TIPC_DIRECT_MSG);
1358 msg_set_lookup_scope(hdr, 0);
1359 msg_set_destnode(hdr, dnode);
1360 msg_set_destport(hdr, dest->addr.id.ref);
1361 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1366 /* Block or return if destination link is congested */
1367 rc = tipc_wait_for_cond(sock, &timeout,
1368 !tipc_dest_find(clinks, dnode, 0));
1372 skb_queue_head_init(&pkts);
1373 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1374 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1375 if (unlikely(rc != dlen))
1377 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1380 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1381 if (unlikely(rc == -ELINKCONG)) {
1382 tipc_dest_push(clinks, dnode, 0);
1383 tsk->cong_link_cnt++;
1387 if (unlikely(syn && !rc))
1388 tipc_set_sk_state(sk, TIPC_CONNECTING);
1390 return rc ? rc : dlen;
1394 * tipc_sendstream - send stream-oriented data
1395 * @sock: socket structure
1397 * @dsz: total length of data to be transmitted
1399 * Used for SOCK_STREAM data.
1401 * Returns the number of bytes sent on success (or partial success),
1402 * or errno if no data sent
1404 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1406 struct sock *sk = sock->sk;
1410 ret = __tipc_sendstream(sock, m, dsz);
1416 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1418 struct sock *sk = sock->sk;
1419 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1420 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1421 struct tipc_sock *tsk = tipc_sk(sk);
1422 struct tipc_msg *hdr = &tsk->phdr;
1423 struct net *net = sock_net(sk);
1424 struct sk_buff_head pkts;
1425 u32 dnode = tsk_peer_node(tsk);
1429 skb_queue_head_init(&pkts);
1431 if (unlikely(dlen > INT_MAX))
1434 /* Handle implicit connection setup */
1435 if (unlikely(dest)) {
1436 rc = __tipc_sendmsg(sock, m, dlen);
1437 if (dlen && dlen == rc) {
1438 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1439 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1445 rc = tipc_wait_for_cond(sock, &timeout,
1446 (!tsk->cong_link_cnt &&
1447 !tsk_conn_cong(tsk) &&
1448 tipc_sk_connected(sk)));
1452 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1453 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1454 if (unlikely(rc != send))
1457 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1458 if (unlikely(rc == -ELINKCONG)) {
1459 tsk->cong_link_cnt = 1;
1463 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1466 } while (sent < dlen && !rc);
1468 return sent ? sent : rc;
1472 * tipc_send_packet - send a connection-oriented message
1473 * @sock: socket structure
1474 * @m: message to send
1475 * @dsz: length of data to be transmitted
1477 * Used for SOCK_SEQPACKET messages.
1479 * Returns the number of bytes sent on success, or errno otherwise
1481 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1483 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1486 return tipc_sendstream(sock, m, dsz);
1489 /* tipc_sk_finish_conn - complete the setup of a connection
1491 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1494 struct sock *sk = &tsk->sk;
1495 struct net *net = sock_net(sk);
1496 struct tipc_msg *msg = &tsk->phdr;
1498 msg_set_syn(msg, 0);
1499 msg_set_destnode(msg, peer_node);
1500 msg_set_destport(msg, peer_port);
1501 msg_set_type(msg, TIPC_CONN_MSG);
1502 msg_set_lookup_scope(msg, 0);
1503 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1505 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1506 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1507 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1508 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1509 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1510 __skb_queue_purge(&sk->sk_write_queue);
1511 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1514 /* Fall back to message based flow control */
1515 tsk->rcv_win = FLOWCTL_MSG_WIN;
1516 tsk->snd_win = FLOWCTL_MSG_WIN;
1520 * tipc_sk_set_orig_addr - capture sender's address for received message
1521 * @m: descriptor for message info
1522 * @hdr: received message header
1524 * Note: Address is not captured if not requested by receiver.
1526 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1528 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1529 struct tipc_msg *hdr = buf_msg(skb);
1534 srcaddr->sock.family = AF_TIPC;
1535 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1536 srcaddr->sock.scope = 0;
1537 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1538 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1539 srcaddr->sock.addr.name.domain = 0;
1540 m->msg_namelen = sizeof(struct sockaddr_tipc);
1542 if (!msg_in_group(hdr))
1545 /* Group message users may also want to know sending member's id */
1546 srcaddr->member.family = AF_TIPC;
1547 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1548 srcaddr->member.scope = 0;
1549 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1550 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1551 srcaddr->member.addr.name.domain = 0;
1552 m->msg_namelen = sizeof(*srcaddr);
1556 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1557 * @m: descriptor for message info
1558 * @msg: received message header
1559 * @tsk: TIPC port associated with message
1561 * Note: Ancillary data is not captured if not requested by receiver.
1563 * Returns 0 if successful, otherwise errno
1565 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1566 struct tipc_sock *tsk)
1574 if (likely(m->msg_controllen == 0))
1577 /* Optionally capture errored message object(s) */
1578 err = msg ? msg_errcode(msg) : 0;
1579 if (unlikely(err)) {
1581 anc_data[1] = msg_data_sz(msg);
1582 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1586 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1593 /* Optionally capture message destination object */
1594 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1595 switch (dest_type) {
1596 case TIPC_NAMED_MSG:
1598 anc_data[0] = msg_nametype(msg);
1599 anc_data[1] = msg_namelower(msg);
1600 anc_data[2] = msg_namelower(msg);
1602 case TIPC_MCAST_MSG:
1604 anc_data[0] = msg_nametype(msg);
1605 anc_data[1] = msg_namelower(msg);
1606 anc_data[2] = msg_nameupper(msg);
1609 has_name = (tsk->conn_type != 0);
1610 anc_data[0] = tsk->conn_type;
1611 anc_data[1] = tsk->conn_instance;
1612 anc_data[2] = tsk->conn_instance;
1618 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1626 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1628 struct sock *sk = &tsk->sk;
1629 struct net *net = sock_net(sk);
1630 struct sk_buff *skb = NULL;
1631 struct tipc_msg *msg;
1632 u32 peer_port = tsk_peer_port(tsk);
1633 u32 dnode = tsk_peer_node(tsk);
1635 if (!tipc_sk_connected(sk))
1637 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1638 dnode, tsk_own_node(tsk), peer_port,
1639 tsk->portid, TIPC_OK);
1643 msg_set_conn_ack(msg, tsk->rcv_unacked);
1644 tsk->rcv_unacked = 0;
1646 /* Adjust to and advertize the correct window limit */
1647 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1648 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1649 msg_set_adv_win(msg, tsk->rcv_win);
1651 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1654 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1656 struct sock *sk = sock->sk;
1658 long timeo = *timeop;
1659 int err = sock_error(sk);
1665 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1666 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1667 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1672 timeo = schedule_timeout(timeo);
1676 if (!skb_queue_empty(&sk->sk_receive_queue))
1681 err = sock_intr_errno(timeo);
1682 if (signal_pending(current))
1685 err = sock_error(sk);
1689 finish_wait(sk_sleep(sk), &wait);
1695 * tipc_recvmsg - receive packet-oriented message
1696 * @m: descriptor for message info
1697 * @buflen: length of user buffer area
1698 * @flags: receive flags
1700 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1701 * If the complete message doesn't fit in user area, truncate it.
1703 * Returns size of returned message data, errno otherwise
1705 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1706 size_t buflen, int flags)
1708 struct sock *sk = sock->sk;
1709 bool connected = !tipc_sk_type_connectionless(sk);
1710 struct tipc_sock *tsk = tipc_sk(sk);
1711 int rc, err, hlen, dlen, copy;
1712 struct sk_buff_head xmitq;
1713 struct tipc_msg *hdr;
1714 struct sk_buff *skb;
1718 /* Catch invalid receive requests */
1719 if (unlikely(!buflen))
1723 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1727 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1729 /* Step rcv queue to first msg with data or error; wait if necessary */
1731 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1734 skb = skb_peek(&sk->sk_receive_queue);
1736 dlen = msg_data_sz(hdr);
1737 hlen = msg_hdr_sz(hdr);
1738 err = msg_errcode(hdr);
1739 grp_evt = msg_is_grp_evt(hdr);
1740 if (likely(dlen || err))
1742 tsk_advance_rx_queue(sk);
1745 /* Collect msg meta data, including error code and rejected data */
1746 tipc_sk_set_orig_addr(m, skb);
1747 rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1751 /* Capture data if non-error msg, otherwise just set return value */
1753 copy = min_t(int, dlen, buflen);
1754 if (unlikely(copy != dlen))
1755 m->msg_flags |= MSG_TRUNC;
1756 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1760 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1766 /* Mark message as group event if applicable */
1767 if (unlikely(grp_evt)) {
1768 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1769 m->msg_flags |= MSG_EOR;
1770 m->msg_flags |= MSG_OOB;
1774 /* Caption of data or error code/rejected data was successful */
1775 if (unlikely(flags & MSG_PEEK))
1778 /* Send group flow control advertisement when applicable */
1779 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1780 skb_queue_head_init(&xmitq);
1781 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1782 msg_orignode(hdr), msg_origport(hdr),
1784 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1787 tsk_advance_rx_queue(sk);
1789 if (likely(!connected))
1792 /* Send connection flow control advertisement when applicable */
1793 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1794 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1795 tipc_sk_send_ack(tsk);
1798 return rc ? rc : copy;
1802 * tipc_recvstream - receive stream-oriented data
1803 * @m: descriptor for message info
1804 * @buflen: total size of user buffer area
1805 * @flags: receive flags
1807 * Used for SOCK_STREAM messages only. If not enough data is available
1808 * will optionally wait for more; never truncates data.
1810 * Returns size of returned message data, errno otherwise
1812 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1813 size_t buflen, int flags)
1815 struct sock *sk = sock->sk;
1816 struct tipc_sock *tsk = tipc_sk(sk);
1817 struct sk_buff *skb;
1818 struct tipc_msg *hdr;
1819 struct tipc_skb_cb *skb_cb;
1820 bool peek = flags & MSG_PEEK;
1821 int offset, required, copy, copied = 0;
1822 int hlen, dlen, err, rc;
1825 /* Catch invalid receive attempts */
1826 if (unlikely(!buflen))
1831 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1835 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1836 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1839 /* Look at first msg in receive queue; wait if necessary */
1840 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1843 skb = skb_peek(&sk->sk_receive_queue);
1844 skb_cb = TIPC_SKB_CB(skb);
1846 dlen = msg_data_sz(hdr);
1847 hlen = msg_hdr_sz(hdr);
1848 err = msg_errcode(hdr);
1850 /* Discard any empty non-errored (SYN-) message */
1851 if (unlikely(!dlen && !err)) {
1852 tsk_advance_rx_queue(sk);
1856 /* Collect msg meta data, incl. error code and rejected data */
1858 tipc_sk_set_orig_addr(m, skb);
1859 rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1864 /* Copy data if msg ok, otherwise return error/partial data */
1866 offset = skb_cb->bytes_read;
1867 copy = min_t(int, dlen - offset, buflen - copied);
1868 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1873 if (unlikely(offset < dlen)) {
1875 skb_cb->bytes_read = offset;
1880 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1889 tsk_advance_rx_queue(sk);
1891 /* Send connection flow control advertisement when applicable */
1892 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1893 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1894 tipc_sk_send_ack(tsk);
1896 /* Exit if all requested data or FIN/error received */
1897 if (copied == buflen || err)
1900 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1903 return copied ? copied : rc;
1907 * tipc_write_space - wake up thread if port congestion is released
1910 static void tipc_write_space(struct sock *sk)
1912 struct socket_wq *wq;
1915 wq = rcu_dereference(sk->sk_wq);
1916 if (skwq_has_sleeper(wq))
1917 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1918 EPOLLWRNORM | EPOLLWRBAND);
1923 * tipc_data_ready - wake up threads to indicate messages have been received
1925 * @len: the length of messages
1927 static void tipc_data_ready(struct sock *sk)
1929 struct socket_wq *wq;
1932 wq = rcu_dereference(sk->sk_wq);
1933 if (skwq_has_sleeper(wq))
1934 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1935 EPOLLRDNORM | EPOLLRDBAND);
1939 static void tipc_sock_destruct(struct sock *sk)
1941 __skb_queue_purge(&sk->sk_receive_queue);
1944 static void tipc_sk_proto_rcv(struct sock *sk,
1945 struct sk_buff_head *inputq,
1946 struct sk_buff_head *xmitq)
1948 struct sk_buff *skb = __skb_dequeue(inputq);
1949 struct tipc_sock *tsk = tipc_sk(sk);
1950 struct tipc_msg *hdr = buf_msg(skb);
1951 struct tipc_group *grp = tsk->group;
1952 bool wakeup = false;
1954 switch (msg_user(hdr)) {
1956 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1959 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1960 tsk->cong_link_cnt--;
1963 case GROUP_PROTOCOL:
1964 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1967 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1968 hdr, inputq, xmitq);
1975 sk->sk_write_space(sk);
1981 * tipc_sk_filter_connect - check incoming message for a connection-based socket
1983 * @skb: pointer to message buffer.
1984 * Returns true if message should be added to receive queue, false otherwise
1986 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1988 struct sock *sk = &tsk->sk;
1989 struct net *net = sock_net(sk);
1990 struct tipc_msg *hdr = buf_msg(skb);
1991 bool con_msg = msg_connected(hdr);
1992 u32 pport = tsk_peer_port(tsk);
1993 u32 pnode = tsk_peer_node(tsk);
1994 u32 oport = msg_origport(hdr);
1995 u32 onode = msg_orignode(hdr);
1996 int err = msg_errcode(hdr);
1997 unsigned long delay;
1999 if (unlikely(msg_mcast(hdr)))
2002 switch (sk->sk_state) {
2003 case TIPC_CONNECTING:
2005 if (likely(con_msg)) {
2008 tipc_sk_finish_conn(tsk, oport, onode);
2009 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2010 /* ACK+ message with data is added to receive queue */
2011 if (msg_data_sz(hdr))
2013 /* Empty ACK-, - wake up sleeping connect() and drop */
2014 sk->sk_data_ready(sk);
2015 msg_set_dest_droppable(hdr, 1);
2018 /* Ignore connectionless message if not from listening socket */
2019 if (oport != pport || onode != pnode)
2023 if (err != TIPC_ERR_OVERLOAD)
2026 /* Prepare for new setup attempt if we have a SYN clone */
2027 if (skb_queue_empty(&sk->sk_write_queue))
2029 get_random_bytes(&delay, 2);
2030 delay %= (tsk->conn_timeout / 4);
2031 delay = msecs_to_jiffies(delay + 100);
2032 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2035 case TIPC_DISCONNECTING:
2038 /* Accept only SYN message */
2039 if (!msg_is_syn(hdr) &&
2040 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2042 if (!con_msg && !err)
2045 case TIPC_ESTABLISHED:
2046 /* Accept only connection-based messages sent by peer */
2047 if (likely(con_msg && !err && pport == oport && pnode == onode))
2049 if (!tsk_peer_msg(tsk, hdr))
2053 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2054 tipc_node_remove_conn(net, pnode, tsk->portid);
2055 sk->sk_state_change(sk);
2058 pr_err("Unknown sk_state %u\n", sk->sk_state);
2060 /* Abort connection setup attempt */
2061 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2062 sk->sk_err = ECONNREFUSED;
2063 sk->sk_state_change(sk);
2068 * rcvbuf_limit - get proper overload limit of socket receive queue
2072 * For connection oriented messages, irrespective of importance,
2073 * default queue limit is 2 MB.
2075 * For connectionless messages, queue limits are based on message
2076 * importance as follows:
2078 * TIPC_LOW_IMPORTANCE (2 MB)
2079 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2080 * TIPC_HIGH_IMPORTANCE (8 MB)
2081 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2083 * Returns overload limit according to corresponding message importance
2085 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2087 struct tipc_sock *tsk = tipc_sk(sk);
2088 struct tipc_msg *hdr = buf_msg(skb);
2090 if (unlikely(msg_in_group(hdr)))
2091 return sk->sk_rcvbuf;
2093 if (unlikely(!msg_connected(hdr)))
2094 return sk->sk_rcvbuf << msg_importance(hdr);
2096 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2097 return sk->sk_rcvbuf;
2099 return FLOWCTL_MSG_LIM;
2103 * tipc_sk_filter_rcv - validate incoming message
2105 * @skb: pointer to message.
2107 * Enqueues message on receive queue if acceptable; optionally handles
2108 * disconnect indication for a connected socket.
2110 * Called with socket lock already taken
2113 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2114 struct sk_buff_head *xmitq)
2116 bool sk_conn = !tipc_sk_type_connectionless(sk);
2117 struct tipc_sock *tsk = tipc_sk(sk);
2118 struct tipc_group *grp = tsk->group;
2119 struct tipc_msg *hdr = buf_msg(skb);
2120 struct net *net = sock_net(sk);
2121 struct sk_buff_head inputq;
2122 int limit, err = TIPC_OK;
2124 TIPC_SKB_CB(skb)->bytes_read = 0;
2125 __skb_queue_head_init(&inputq);
2126 __skb_queue_tail(&inputq, skb);
2128 if (unlikely(!msg_isdata(hdr)))
2129 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2132 tipc_group_filter_msg(grp, &inputq, xmitq);
2134 /* Validate and add to receive buffer if there is space */
2135 while ((skb = __skb_dequeue(&inputq))) {
2137 limit = rcvbuf_limit(sk, skb);
2138 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2139 (!sk_conn && msg_connected(hdr)) ||
2140 (!grp && msg_in_group(hdr)))
2141 err = TIPC_ERR_NO_PORT;
2142 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2143 atomic_inc(&sk->sk_drops);
2144 err = TIPC_ERR_OVERLOAD;
2147 if (unlikely(err)) {
2148 tipc_skb_reject(net, err, skb, xmitq);
2152 __skb_queue_tail(&sk->sk_receive_queue, skb);
2153 skb_set_owner_r(skb, sk);
2154 sk->sk_data_ready(sk);
2159 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2163 * Caller must hold socket lock
2165 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2167 unsigned int before = sk_rmem_alloc_get(sk);
2168 struct sk_buff_head xmitq;
2171 __skb_queue_head_init(&xmitq);
2173 tipc_sk_filter_rcv(sk, skb, &xmitq);
2174 added = sk_rmem_alloc_get(sk) - before;
2175 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2177 /* Send pending response/rejected messages, if any */
2178 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2183 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2184 * inputq and try adding them to socket or backlog queue
2185 * @inputq: list of incoming buffers with potentially different destinations
2186 * @sk: socket where the buffers should be enqueued
2187 * @dport: port number for the socket
2189 * Caller must hold socket lock
2191 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2192 u32 dport, struct sk_buff_head *xmitq)
2194 unsigned long time_limit = jiffies + 2;
2195 struct sk_buff *skb;
2200 while (skb_queue_len(inputq)) {
2201 if (unlikely(time_after_eq(jiffies, time_limit)))
2204 skb = tipc_skb_dequeue(inputq, dport);
2208 /* Add message directly to receive queue if possible */
2209 if (!sock_owned_by_user(sk)) {
2210 tipc_sk_filter_rcv(sk, skb, xmitq);
2214 /* Try backlog, compensating for double-counted bytes */
2215 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2216 if (!sk->sk_backlog.len)
2217 atomic_set(dcnt, 0);
2218 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2219 if (likely(!sk_add_backlog(sk, skb, lim)))
2222 /* Overload => reject message back to sender */
2223 onode = tipc_own_addr(sock_net(sk));
2224 atomic_inc(&sk->sk_drops);
2225 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2226 __skb_queue_tail(xmitq, skb);
2232 * tipc_sk_rcv - handle a chain of incoming buffers
2233 * @inputq: buffer list containing the buffers
2234 * Consumes all buffers in list until inputq is empty
2235 * Note: may be called in multiple threads referring to the same queue
2237 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2239 struct sk_buff_head xmitq;
2240 u32 dnode, dport = 0;
2242 struct tipc_sock *tsk;
2244 struct sk_buff *skb;
2246 __skb_queue_head_init(&xmitq);
2247 while (skb_queue_len(inputq)) {
2248 dport = tipc_skb_peek_port(inputq, dport);
2249 tsk = tipc_sk_lookup(net, dport);
2253 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2254 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2255 spin_unlock_bh(&sk->sk_lock.slock);
2257 /* Send pending response/rejected messages, if any */
2258 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2262 /* No destination socket => dequeue skb if still there */
2263 skb = tipc_skb_dequeue(inputq, dport);
2267 /* Try secondary lookup if unresolved named message */
2268 err = TIPC_ERR_NO_PORT;
2269 if (tipc_msg_lookup_dest(net, skb, &err))
2272 /* Prepare for message rejection */
2273 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2276 dnode = msg_destnode(buf_msg(skb));
2277 tipc_node_xmit_skb(net, skb, dnode, dport);
2281 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2283 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2284 struct sock *sk = sock->sk;
2288 int err = sock_error(sk);
2293 if (signal_pending(current))
2294 return sock_intr_errno(*timeo_p);
2296 add_wait_queue(sk_sleep(sk), &wait);
2297 done = sk_wait_event(sk, timeo_p,
2298 sk->sk_state != TIPC_CONNECTING, &wait);
2299 remove_wait_queue(sk_sleep(sk), &wait);
2305 * tipc_connect - establish a connection to another TIPC port
2306 * @sock: socket structure
2307 * @dest: socket address for destination port
2308 * @destlen: size of socket address data structure
2309 * @flags: file-related flags associated with socket
2311 * Returns 0 on success, errno otherwise
2313 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2314 int destlen, int flags)
2316 struct sock *sk = sock->sk;
2317 struct tipc_sock *tsk = tipc_sk(sk);
2318 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2319 struct msghdr m = {NULL,};
2320 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2324 if (destlen != sizeof(struct sockaddr_tipc))
2334 if (dst->family == AF_UNSPEC) {
2335 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2336 if (!tipc_sk_type_connectionless(sk))
2339 } else if (dst->family != AF_TIPC) {
2342 if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2347 /* DGRAM/RDM connect(), just save the destaddr */
2348 if (tipc_sk_type_connectionless(sk)) {
2349 memcpy(&tsk->peer, dest, destlen);
2353 previous = sk->sk_state;
2355 switch (sk->sk_state) {
2357 /* Send a 'SYN-' to destination */
2359 m.msg_namelen = destlen;
2361 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2362 * indicate send_msg() is never blocked.
2365 m.msg_flags = MSG_DONTWAIT;
2367 res = __tipc_sendmsg(sock, &m, 0);
2368 if ((res < 0) && (res != -EWOULDBLOCK))
2371 /* Just entered TIPC_CONNECTING state; the only
2372 * difference is that return value in non-blocking
2373 * case is EINPROGRESS, rather than EALREADY.
2377 case TIPC_CONNECTING:
2379 if (previous == TIPC_CONNECTING)
2383 timeout = msecs_to_jiffies(timeout);
2384 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2385 res = tipc_wait_for_connect(sock, &timeout);
2387 case TIPC_ESTABLISHED:
2400 * tipc_listen - allow socket to listen for incoming connections
2401 * @sock: socket structure
2404 * Returns 0 on success, errno otherwise
2406 static int tipc_listen(struct socket *sock, int len)
2408 struct sock *sk = sock->sk;
2412 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2418 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2420 struct sock *sk = sock->sk;
2424 /* True wake-one mechanism for incoming connections: only
2425 * one process gets woken up, not the 'whole herd'.
2426 * Since we do not 'race & poll' for established sockets
2427 * anymore, the common case will execute the loop only once.
2430 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2431 TASK_INTERRUPTIBLE);
2432 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2434 timeo = schedule_timeout(timeo);
2438 if (!skb_queue_empty(&sk->sk_receive_queue))
2443 err = sock_intr_errno(timeo);
2444 if (signal_pending(current))
2447 finish_wait(sk_sleep(sk), &wait);
2452 * tipc_accept - wait for connection request
2453 * @sock: listening socket
2454 * @newsock: new socket that is to be connected
2455 * @flags: file-related flags associated with socket
2457 * Returns 0 on success, errno otherwise
2459 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2462 struct sock *new_sk, *sk = sock->sk;
2463 struct sk_buff *buf;
2464 struct tipc_sock *new_tsock;
2465 struct tipc_msg *msg;
2471 if (sk->sk_state != TIPC_LISTEN) {
2475 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2476 res = tipc_wait_for_accept(sock, timeo);
2480 buf = skb_peek(&sk->sk_receive_queue);
2482 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2485 security_sk_clone(sock->sk, new_sock->sk);
2487 new_sk = new_sock->sk;
2488 new_tsock = tipc_sk(new_sk);
2491 /* we lock on new_sk; but lockdep sees the lock on sk */
2492 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2495 * Reject any stray messages received by new socket
2496 * before the socket lock was taken (very, very unlikely)
2498 tsk_rej_rx_queue(new_sk);
2500 /* Connect new socket to it's peer */
2501 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2503 tsk_set_importance(new_tsock, msg_importance(msg));
2504 if (msg_named(msg)) {
2505 new_tsock->conn_type = msg_nametype(msg);
2506 new_tsock->conn_instance = msg_nameinst(msg);
2510 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2511 * Respond to 'SYN+' by queuing it on new socket.
2513 if (!msg_data_sz(msg)) {
2514 struct msghdr m = {NULL,};
2516 tsk_advance_rx_queue(sk);
2517 __tipc_sendstream(new_sock, &m, 0);
2519 __skb_dequeue(&sk->sk_receive_queue);
2520 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2521 skb_set_owner_r(buf, new_sk);
2523 release_sock(new_sk);
2530 * tipc_shutdown - shutdown socket connection
2531 * @sock: socket structure
2532 * @how: direction to close (must be SHUT_RDWR)
2534 * Terminates connection (if necessary), then purges socket's receive queue.
2536 * Returns 0 on success, errno otherwise
2538 static int tipc_shutdown(struct socket *sock, int how)
2540 struct sock *sk = sock->sk;
2543 if (how != SHUT_RDWR)
2548 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2549 sk->sk_shutdown = SEND_SHUTDOWN;
2551 if (sk->sk_state == TIPC_DISCONNECTING) {
2552 /* Discard any unreceived messages */
2553 __skb_queue_purge(&sk->sk_receive_queue);
2555 /* Wake up anyone sleeping in poll */
2556 sk->sk_state_change(sk);
2566 static void tipc_sk_check_probing_state(struct sock *sk,
2567 struct sk_buff_head *list)
2569 struct tipc_sock *tsk = tipc_sk(sk);
2570 u32 pnode = tsk_peer_node(tsk);
2571 u32 pport = tsk_peer_port(tsk);
2572 u32 self = tsk_own_node(tsk);
2573 u32 oport = tsk->portid;
2574 struct sk_buff *skb;
2576 if (tsk->probe_unacked) {
2577 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2578 sk->sk_err = ECONNABORTED;
2579 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2580 sk->sk_state_change(sk);
2583 /* Prepare new probe */
2584 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2585 pnode, self, pport, oport, TIPC_OK);
2587 __skb_queue_tail(list, skb);
2588 tsk->probe_unacked = true;
2589 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2592 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2594 struct tipc_sock *tsk = tipc_sk(sk);
2596 /* Try again later if dest link is congested */
2597 if (tsk->cong_link_cnt) {
2598 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2601 /* Prepare SYN for retransmit */
2602 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2605 static void tipc_sk_timeout(struct timer_list *t)
2607 struct sock *sk = from_timer(sk, t, sk_timer);
2608 struct tipc_sock *tsk = tipc_sk(sk);
2609 u32 pnode = tsk_peer_node(tsk);
2610 struct sk_buff_head list;
2613 skb_queue_head_init(&list);
2616 /* Try again later if socket is busy */
2617 if (sock_owned_by_user(sk)) {
2618 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2623 if (sk->sk_state == TIPC_ESTABLISHED)
2624 tipc_sk_check_probing_state(sk, &list);
2625 else if (sk->sk_state == TIPC_CONNECTING)
2626 tipc_sk_retry_connect(sk, &list);
2630 if (!skb_queue_empty(&list))
2631 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2633 /* SYN messages may cause link congestion */
2634 if (rc == -ELINKCONG) {
2635 tipc_dest_push(&tsk->cong_links, pnode, 0);
2636 tsk->cong_link_cnt = 1;
2641 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2642 struct tipc_name_seq const *seq)
2644 struct sock *sk = &tsk->sk;
2645 struct net *net = sock_net(sk);
2646 struct publication *publ;
2649 if (scope != TIPC_NODE_SCOPE)
2650 scope = TIPC_CLUSTER_SCOPE;
2652 if (tipc_sk_connected(sk))
2654 key = tsk->portid + tsk->pub_count + 1;
2655 if (key == tsk->portid)
2658 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2659 scope, tsk->portid, key);
2660 if (unlikely(!publ))
2663 list_add(&publ->binding_sock, &tsk->publications);
2669 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2670 struct tipc_name_seq const *seq)
2672 struct net *net = sock_net(&tsk->sk);
2673 struct publication *publ;
2674 struct publication *safe;
2677 if (scope != TIPC_NODE_SCOPE)
2678 scope = TIPC_CLUSTER_SCOPE;
2680 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2682 if (publ->scope != scope)
2684 if (publ->type != seq->type)
2686 if (publ->lower != seq->lower)
2688 if (publ->upper != seq->upper)
2690 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2691 publ->upper, publ->key);
2695 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2696 publ->upper, publ->key);
2699 if (list_empty(&tsk->publications))
2704 /* tipc_sk_reinit: set non-zero address in all existing sockets
2705 * when we go from standalone to network mode.
2707 void tipc_sk_reinit(struct net *net)
2709 struct tipc_net *tn = net_generic(net, tipc_net_id);
2710 struct rhashtable_iter iter;
2711 struct tipc_sock *tsk;
2712 struct tipc_msg *msg;
2714 rhashtable_walk_enter(&tn->sk_rht, &iter);
2717 rhashtable_walk_start(&iter);
2719 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2720 spin_lock_bh(&tsk->sk.sk_lock.slock);
2722 msg_set_prevnode(msg, tipc_own_addr(net));
2723 msg_set_orignode(msg, tipc_own_addr(net));
2724 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2727 rhashtable_walk_stop(&iter);
2728 } while (tsk == ERR_PTR(-EAGAIN));
2730 rhashtable_walk_exit(&iter);
2733 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2735 struct tipc_net *tn = net_generic(net, tipc_net_id);
2736 struct tipc_sock *tsk;
2739 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2741 sock_hold(&tsk->sk);
2747 static int tipc_sk_insert(struct tipc_sock *tsk)
2749 struct sock *sk = &tsk->sk;
2750 struct net *net = sock_net(sk);
2751 struct tipc_net *tn = net_generic(net, tipc_net_id);
2752 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2753 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2755 while (remaining--) {
2757 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2758 portid = TIPC_MIN_PORT;
2759 tsk->portid = portid;
2760 sock_hold(&tsk->sk);
2761 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2770 static void tipc_sk_remove(struct tipc_sock *tsk)
2772 struct sock *sk = &tsk->sk;
2773 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2775 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2776 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2781 static const struct rhashtable_params tsk_rht_params = {
2783 .head_offset = offsetof(struct tipc_sock, node),
2784 .key_offset = offsetof(struct tipc_sock, portid),
2785 .key_len = sizeof(u32), /* portid */
2786 .max_size = 1048576,
2788 .automatic_shrinking = true,
2791 int tipc_sk_rht_init(struct net *net)
2793 struct tipc_net *tn = net_generic(net, tipc_net_id);
2795 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2798 void tipc_sk_rht_destroy(struct net *net)
2800 struct tipc_net *tn = net_generic(net, tipc_net_id);
2802 /* Wait for socket readers to complete */
2805 rhashtable_destroy(&tn->sk_rht);
2808 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2810 struct net *net = sock_net(&tsk->sk);
2811 struct tipc_group *grp = tsk->group;
2812 struct tipc_msg *hdr = &tsk->phdr;
2813 struct tipc_name_seq seq;
2816 if (mreq->type < TIPC_RESERVED_TYPES)
2818 if (mreq->scope > TIPC_NODE_SCOPE)
2822 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2826 msg_set_lookup_scope(hdr, mreq->scope);
2827 msg_set_nametype(hdr, mreq->type);
2828 msg_set_dest_droppable(hdr, true);
2829 seq.type = mreq->type;
2830 seq.lower = mreq->instance;
2831 seq.upper = seq.lower;
2832 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2833 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2835 tipc_group_delete(net, grp);
2839 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2840 tsk->mc_method.rcast = true;
2841 tsk->mc_method.mandatory = true;
2842 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2846 static int tipc_sk_leave(struct tipc_sock *tsk)
2848 struct net *net = sock_net(&tsk->sk);
2849 struct tipc_group *grp = tsk->group;
2850 struct tipc_name_seq seq;
2855 tipc_group_self(grp, &seq, &scope);
2856 tipc_group_delete(net, grp);
2858 tipc_sk_withdraw(tsk, scope, &seq);
2863 * tipc_setsockopt - set socket option
2864 * @sock: socket structure
2865 * @lvl: option level
2866 * @opt: option identifier
2867 * @ov: pointer to new option value
2868 * @ol: length of option value
2870 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2871 * (to ease compatibility).
2873 * Returns 0 on success, errno otherwise
2875 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2876 char __user *ov, unsigned int ol)
2878 struct sock *sk = sock->sk;
2879 struct tipc_sock *tsk = tipc_sk(sk);
2880 struct tipc_group_req mreq;
2884 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2886 if (lvl != SOL_TIPC)
2887 return -ENOPROTOOPT;
2890 case TIPC_IMPORTANCE:
2891 case TIPC_SRC_DROPPABLE:
2892 case TIPC_DEST_DROPPABLE:
2893 case TIPC_CONN_TIMEOUT:
2894 if (ol < sizeof(value))
2896 if (get_user(value, (u32 __user *)ov))
2899 case TIPC_GROUP_JOIN:
2900 if (ol < sizeof(mreq))
2902 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2913 case TIPC_IMPORTANCE:
2914 res = tsk_set_importance(tsk, value);
2916 case TIPC_SRC_DROPPABLE:
2917 if (sock->type != SOCK_STREAM)
2918 tsk_set_unreliable(tsk, value);
2922 case TIPC_DEST_DROPPABLE:
2923 tsk_set_unreturnable(tsk, value);
2925 case TIPC_CONN_TIMEOUT:
2926 tipc_sk(sk)->conn_timeout = value;
2928 case TIPC_MCAST_BROADCAST:
2929 tsk->mc_method.rcast = false;
2930 tsk->mc_method.mandatory = true;
2932 case TIPC_MCAST_REPLICAST:
2933 tsk->mc_method.rcast = true;
2934 tsk->mc_method.mandatory = true;
2936 case TIPC_GROUP_JOIN:
2937 res = tipc_sk_join(tsk, &mreq);
2939 case TIPC_GROUP_LEAVE:
2940 res = tipc_sk_leave(tsk);
2952 * tipc_getsockopt - get socket option
2953 * @sock: socket structure
2954 * @lvl: option level
2955 * @opt: option identifier
2956 * @ov: receptacle for option value
2957 * @ol: receptacle for length of option value
2959 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2960 * (to ease compatibility).
2962 * Returns 0 on success, errno otherwise
2964 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2965 char __user *ov, int __user *ol)
2967 struct sock *sk = sock->sk;
2968 struct tipc_sock *tsk = tipc_sk(sk);
2969 struct tipc_name_seq seq;
2974 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2975 return put_user(0, ol);
2976 if (lvl != SOL_TIPC)
2977 return -ENOPROTOOPT;
2978 res = get_user(len, ol);
2985 case TIPC_IMPORTANCE:
2986 value = tsk_importance(tsk);
2988 case TIPC_SRC_DROPPABLE:
2989 value = tsk_unreliable(tsk);
2991 case TIPC_DEST_DROPPABLE:
2992 value = tsk_unreturnable(tsk);
2994 case TIPC_CONN_TIMEOUT:
2995 value = tsk->conn_timeout;
2996 /* no need to set "res", since already 0 at this point */
2998 case TIPC_NODE_RECVQ_DEPTH:
2999 value = 0; /* was tipc_queue_size, now obsolete */
3001 case TIPC_SOCK_RECVQ_DEPTH:
3002 value = skb_queue_len(&sk->sk_receive_queue);
3004 case TIPC_GROUP_JOIN:
3007 tipc_group_self(tsk->group, &seq, &scope);
3017 return res; /* "get" failed */
3019 if (len < sizeof(value))
3022 if (copy_to_user(ov, &value, sizeof(value)))
3025 return put_user(sizeof(value), ol);
3028 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3030 struct net *net = sock_net(sock->sk);
3031 struct tipc_sioc_nodeid_req nr = {0};
3032 struct tipc_sioc_ln_req lnr;
3033 void __user *argp = (void __user *)arg;
3036 case SIOCGETLINKNAME:
3037 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3039 if (!tipc_node_get_linkname(net,
3040 lnr.bearer_id & 0xffff, lnr.peer,
3041 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3042 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3046 return -EADDRNOTAVAIL;
3048 if (copy_from_user(&nr, argp, sizeof(nr)))
3050 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3051 return -EADDRNOTAVAIL;
3052 if (copy_to_user(argp, &nr, sizeof(nr)))
3056 return -ENOIOCTLCMD;
3060 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3062 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3063 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3064 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3066 tsk1->peer.family = AF_TIPC;
3067 tsk1->peer.addrtype = TIPC_ADDR_ID;
3068 tsk1->peer.scope = TIPC_NODE_SCOPE;
3069 tsk1->peer.addr.id.ref = tsk2->portid;
3070 tsk1->peer.addr.id.node = onode;
3071 tsk2->peer.family = AF_TIPC;
3072 tsk2->peer.addrtype = TIPC_ADDR_ID;
3073 tsk2->peer.scope = TIPC_NODE_SCOPE;
3074 tsk2->peer.addr.id.ref = tsk1->portid;
3075 tsk2->peer.addr.id.node = onode;
3077 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3078 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3082 /* Protocol switches for the various types of TIPC sockets */
3084 static const struct proto_ops msg_ops = {
3085 .owner = THIS_MODULE,
3087 .release = tipc_release,
3089 .connect = tipc_connect,
3090 .socketpair = tipc_socketpair,
3091 .accept = sock_no_accept,
3092 .getname = tipc_getname,
3094 .ioctl = tipc_ioctl,
3095 .listen = sock_no_listen,
3096 .shutdown = tipc_shutdown,
3097 .setsockopt = tipc_setsockopt,
3098 .getsockopt = tipc_getsockopt,
3099 .sendmsg = tipc_sendmsg,
3100 .recvmsg = tipc_recvmsg,
3101 .mmap = sock_no_mmap,
3102 .sendpage = sock_no_sendpage
3105 static const struct proto_ops packet_ops = {
3106 .owner = THIS_MODULE,
3108 .release = tipc_release,
3110 .connect = tipc_connect,
3111 .socketpair = tipc_socketpair,
3112 .accept = tipc_accept,
3113 .getname = tipc_getname,
3115 .ioctl = tipc_ioctl,
3116 .listen = tipc_listen,
3117 .shutdown = tipc_shutdown,
3118 .setsockopt = tipc_setsockopt,
3119 .getsockopt = tipc_getsockopt,
3120 .sendmsg = tipc_send_packet,
3121 .recvmsg = tipc_recvmsg,
3122 .mmap = sock_no_mmap,
3123 .sendpage = sock_no_sendpage
3126 static const struct proto_ops stream_ops = {
3127 .owner = THIS_MODULE,
3129 .release = tipc_release,
3131 .connect = tipc_connect,
3132 .socketpair = tipc_socketpair,
3133 .accept = tipc_accept,
3134 .getname = tipc_getname,
3136 .ioctl = tipc_ioctl,
3137 .listen = tipc_listen,
3138 .shutdown = tipc_shutdown,
3139 .setsockopt = tipc_setsockopt,
3140 .getsockopt = tipc_getsockopt,
3141 .sendmsg = tipc_sendstream,
3142 .recvmsg = tipc_recvstream,
3143 .mmap = sock_no_mmap,
3144 .sendpage = sock_no_sendpage
3147 static const struct net_proto_family tipc_family_ops = {
3148 .owner = THIS_MODULE,
3150 .create = tipc_sk_create
3153 static struct proto tipc_proto = {
3155 .owner = THIS_MODULE,
3156 .obj_size = sizeof(struct tipc_sock),
3157 .sysctl_rmem = sysctl_tipc_rmem
3161 * tipc_socket_init - initialize TIPC socket interface
3163 * Returns 0 on success, errno otherwise
3165 int tipc_socket_init(void)
3169 res = proto_register(&tipc_proto, 1);
3171 pr_err("Failed to register TIPC protocol type\n");
3175 res = sock_register(&tipc_family_ops);
3177 pr_err("Failed to register TIPC socket type\n");
3178 proto_unregister(&tipc_proto);
3186 * tipc_socket_stop - stop TIPC socket interface
3188 void tipc_socket_stop(void)
3190 sock_unregister(tipc_family_ops.family);
3191 proto_unregister(&tipc_proto);
3194 /* Caller should hold socket lock for the passed tipc socket. */
3195 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3199 struct nlattr *nest;
3201 peer_node = tsk_peer_node(tsk);
3202 peer_port = tsk_peer_port(tsk);
3204 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3206 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3208 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3211 if (tsk->conn_type != 0) {
3212 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3214 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3216 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3219 nla_nest_end(skb, nest);
3224 nla_nest_cancel(skb, nest);
3229 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3232 struct net *net = sock_net(skb->sk);
3233 struct sock *sk = &tsk->sk;
3235 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3236 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3239 if (tipc_sk_connected(sk)) {
3240 if (__tipc_nl_add_sk_con(skb, tsk))
3242 } else if (!list_empty(&tsk->publications)) {
3243 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3249 /* Caller should hold socket lock for the passed tipc socket. */
3250 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3251 struct tipc_sock *tsk)
3253 struct nlattr *attrs;
3256 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3257 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3261 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3263 goto genlmsg_cancel;
3265 if (__tipc_nl_add_sk_info(skb, tsk))
3266 goto attr_msg_cancel;
3268 nla_nest_end(skb, attrs);
3269 genlmsg_end(skb, hdr);
3274 nla_nest_cancel(skb, attrs);
3276 genlmsg_cancel(skb, hdr);
3281 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3282 int (*skb_handler)(struct sk_buff *skb,
3283 struct netlink_callback *cb,
3284 struct tipc_sock *tsk))
3286 struct rhashtable_iter *iter = (void *)cb->args[4];
3287 struct tipc_sock *tsk;
3290 rhashtable_walk_start(iter);
3291 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3294 if (err == -EAGAIN) {
3301 sock_hold(&tsk->sk);
3302 rhashtable_walk_stop(iter);
3303 lock_sock(&tsk->sk);
3304 err = skb_handler(skb, cb, tsk);
3306 release_sock(&tsk->sk);
3310 release_sock(&tsk->sk);
3311 rhashtable_walk_start(iter);
3314 rhashtable_walk_stop(iter);
3318 EXPORT_SYMBOL(tipc_nl_sk_walk);
3320 int tipc_dump_start(struct netlink_callback *cb)
3322 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3324 EXPORT_SYMBOL(tipc_dump_start);
3326 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3328 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3329 struct rhashtable_iter *iter = (void *)cb->args[4];
3330 struct tipc_net *tn = tipc_net(net);
3333 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3337 cb->args[4] = (long)iter;
3340 rhashtable_walk_enter(&tn->sk_rht, iter);
3344 int tipc_dump_done(struct netlink_callback *cb)
3346 struct rhashtable_iter *hti = (void *)cb->args[4];
3348 rhashtable_walk_exit(hti);
3352 EXPORT_SYMBOL(tipc_dump_done);
3354 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3355 struct tipc_sock *tsk, u32 sk_filter_state,
3356 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3358 struct sock *sk = &tsk->sk;
3359 struct nlattr *attrs;
3360 struct nlattr *stat;
3362 /*filter response w.r.t sk_state*/
3363 if (!(sk_filter_state & (1 << sk->sk_state)))
3366 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3370 if (__tipc_nl_add_sk_info(skb, tsk))
3371 goto attr_msg_cancel;
3373 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3374 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3375 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3376 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3377 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3379 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3380 tipc_diag_gen_cookie(sk),
3382 goto attr_msg_cancel;
3384 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3386 goto attr_msg_cancel;
3388 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3389 skb_queue_len(&sk->sk_receive_queue)) ||
3390 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3391 skb_queue_len(&sk->sk_write_queue)) ||
3392 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3393 atomic_read(&sk->sk_drops)))
3394 goto stat_msg_cancel;
3396 if (tsk->cong_link_cnt &&
3397 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3398 goto stat_msg_cancel;
3400 if (tsk_conn_cong(tsk) &&
3401 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3402 goto stat_msg_cancel;
3404 nla_nest_end(skb, stat);
3407 if (tipc_group_fill_sock_diag(tsk->group, skb))
3408 goto stat_msg_cancel;
3410 nla_nest_end(skb, attrs);
3415 nla_nest_cancel(skb, stat);
3417 nla_nest_cancel(skb, attrs);
3421 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3423 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3425 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3428 /* Caller should hold socket lock for the passed tipc socket. */
3429 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3430 struct netlink_callback *cb,
3431 struct publication *publ)
3434 struct nlattr *attrs;
3436 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3437 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3441 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3443 goto genlmsg_cancel;
3445 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3446 goto attr_msg_cancel;
3447 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3448 goto attr_msg_cancel;
3449 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3450 goto attr_msg_cancel;
3451 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3452 goto attr_msg_cancel;
3454 nla_nest_end(skb, attrs);
3455 genlmsg_end(skb, hdr);
3460 nla_nest_cancel(skb, attrs);
3462 genlmsg_cancel(skb, hdr);
3467 /* Caller should hold socket lock for the passed tipc socket. */
3468 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3469 struct netlink_callback *cb,
3470 struct tipc_sock *tsk, u32 *last_publ)
3473 struct publication *p;
3476 list_for_each_entry(p, &tsk->publications, binding_sock) {
3477 if (p->key == *last_publ)
3480 if (p->key != *last_publ) {
3481 /* We never set seq or call nl_dump_check_consistent()
3482 * this means that setting prev_seq here will cause the
3483 * consistence check to fail in the netlink callback
3484 * handler. Resulting in the last NLMSG_DONE message
3485 * having the NLM_F_DUMP_INTR flag set.
3492 p = list_first_entry(&tsk->publications, struct publication,
3496 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3497 err = __tipc_nl_add_sk_publ(skb, cb, p);
3499 *last_publ = p->key;
3508 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3511 u32 tsk_portid = cb->args[0];
3512 u32 last_publ = cb->args[1];
3513 u32 done = cb->args[2];
3514 struct net *net = sock_net(skb->sk);
3515 struct tipc_sock *tsk;
3518 struct nlattr **attrs;
3519 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3521 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3525 if (!attrs[TIPC_NLA_SOCK])
3528 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3529 attrs[TIPC_NLA_SOCK],
3530 tipc_nl_sock_policy, NULL);
3534 if (!sock[TIPC_NLA_SOCK_REF])
3537 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3543 tsk = tipc_sk_lookup(net, tsk_portid);
3547 lock_sock(&tsk->sk);
3548 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3551 release_sock(&tsk->sk);
3554 cb->args[0] = tsk_portid;
3555 cb->args[1] = last_publ;