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
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86 * @cong_link_cnt: number of congested links
87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
89 * @peer: 'connected' peer for dgram/rdm
90 * @node: hash table node
91 * @mc_method: cookie for use between socket and broadcast layer
92 * @rcu: rcu struct for tipc_sock
101 struct tipc_msg phdr;
102 struct list_head cong_links;
103 struct list_head publications;
106 atomic_t dupl_rcvcnt;
114 struct sockaddr_tipc peer;
115 struct rhash_head node;
116 struct tipc_mc_method mc_method;
118 struct tipc_group *group;
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
147 static u32 tsk_own_node(struct tipc_sock *tsk)
149 return msg_prevnode(&tsk->phdr);
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
154 return msg_destnode(&tsk->phdr);
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
159 return msg_destport(&tsk->phdr);
162 static bool tsk_unreliable(struct tipc_sock *tsk)
164 return msg_src_droppable(&tsk->phdr) != 0;
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
169 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
174 return msg_dest_droppable(&tsk->phdr) != 0;
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
179 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
182 static int tsk_importance(struct tipc_sock *tsk)
184 return msg_importance(&tsk->phdr);
187 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
189 if (imp > TIPC_CRITICAL_IMPORTANCE)
191 msg_set_importance(&tsk->phdr, (u32)imp);
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
197 return container_of(sk, struct tipc_sock, sk);
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
202 return tsk->snt_unacked > tsk->snd_win;
205 static u16 tsk_blocks(int len)
207 return ((len / FLOWCTL_BLK_SZ) + 1);
210 /* tsk_blocks(): translate a buffer size in bytes to number of
211 * advertisable blocks, taking into account the ratio truesize(len)/len
212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
214 static u16 tsk_adv_blocks(int len)
216 return len / FLOWCTL_BLK_SZ / 4;
219 /* tsk_inc(): increment counter for sent or received data
220 * - If block based flow control is not supported by peer we
221 * fall back to message based ditto, incrementing the counter
223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
225 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 return ((msglen / FLOWCTL_BLK_SZ) + 1);
231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
233 * Caller must hold socket lock
235 static void tsk_advance_rx_queue(struct sock *sk)
237 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
240 /* tipc_sk_respond() : send response message back to sender
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
246 u32 onode = tipc_own_addr(sock_net(sk));
248 if (!tipc_msg_reverse(onode, &skb, err))
251 dnode = msg_destnode(buf_msg(skb));
252 selector = msg_origport(buf_msg(skb));
253 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
257 * tsk_rej_rx_queue - reject all buffers in socket receive queue
259 * Caller must hold socket lock
261 static void tsk_rej_rx_queue(struct sock *sk)
265 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
269 static bool tipc_sk_connected(struct sock *sk)
271 return sk->sk_state == TIPC_ESTABLISHED;
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
277 * Returns true if connection less, false otherwise
279 static bool tipc_sk_type_connectionless(struct sock *sk)
281 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
286 * Handles cases where the node's network address has changed from
287 * the default of <0.0.0> to its configured setting.
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
291 struct sock *sk = &tsk->sk;
292 u32 self = tipc_own_addr(sock_net(sk));
293 u32 peer_port = tsk_peer_port(tsk);
294 u32 orig_node, peer_node;
296 if (unlikely(!tipc_sk_connected(sk)))
299 if (unlikely(msg_origport(msg) != peer_port))
302 orig_node = msg_orignode(msg);
303 peer_node = tsk_peer_node(tsk);
305 if (likely(orig_node == peer_node))
308 if (!orig_node && peer_node == self)
311 if (!peer_node && orig_node == self)
317 /* tipc_set_sk_state - set the sk_state of the socket
320 * Caller must hold socket lock
322 * Returns 0 on success, errno otherwise
324 static int tipc_set_sk_state(struct sock *sk, int state)
326 int oldsk_state = sk->sk_state;
334 case TIPC_CONNECTING:
335 if (oldsk_state == TIPC_OPEN)
338 case TIPC_ESTABLISHED:
339 if (oldsk_state == TIPC_CONNECTING ||
340 oldsk_state == TIPC_OPEN)
343 case TIPC_DISCONNECTING:
344 if (oldsk_state == TIPC_CONNECTING ||
345 oldsk_state == TIPC_ESTABLISHED)
351 sk->sk_state = state;
356 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
358 struct sock *sk = sock->sk;
359 int err = sock_error(sk);
360 int typ = sock->type;
364 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
365 if (sk->sk_state == TIPC_DISCONNECTING)
367 else if (!tipc_sk_connected(sk))
372 if (signal_pending(current))
373 return sock_intr_errno(*timeout);
378 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
383 while ((rc_ = !(condition_))) { \
384 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
386 rc_ = tipc_sk_sock_err((sock_), timeo_); \
389 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \
391 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 sched_annotate_sleep(); \
394 remove_wait_queue(sk_sleep(sk_), &wait_); \
400 * tipc_sk_create - create a TIPC socket
401 * @net: network namespace (must be default network)
402 * @sock: pre-allocated socket structure
403 * @protocol: protocol indicator (must be 0)
404 * @kern: caused by kernel or by userspace?
406 * This routine creates additional data structures used by the TIPC socket,
407 * initializes them, and links them together.
409 * Returns 0 on success, errno otherwise
411 static int tipc_sk_create(struct net *net, struct socket *sock,
412 int protocol, int kern)
414 const struct proto_ops *ops;
416 struct tipc_sock *tsk;
417 struct tipc_msg *msg;
419 /* Validate arguments */
420 if (unlikely(protocol != 0))
421 return -EPROTONOSUPPORT;
423 switch (sock->type) {
438 /* Allocate socket's protocol area */
439 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
444 tsk->max_pkt = MAX_PKT_DEFAULT;
445 INIT_LIST_HEAD(&tsk->publications);
446 INIT_LIST_HEAD(&tsk->cong_links);
449 /* Finish initializing socket data structures */
451 sock_init_data(sock, sk);
452 tipc_set_sk_state(sk, TIPC_OPEN);
453 if (tipc_sk_insert(tsk)) {
454 pr_warn("Socket create failed; port number exhausted\n");
458 /* Ensure tsk is visible before we read own_addr. */
461 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
462 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
464 msg_set_origport(msg, tsk->portid);
465 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
467 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
468 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
469 sk->sk_data_ready = tipc_data_ready;
470 sk->sk_write_space = tipc_write_space;
471 sk->sk_destruct = tipc_sock_destruct;
472 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
473 tsk->group_is_open = true;
474 atomic_set(&tsk->dupl_rcvcnt, 0);
476 /* Start out with safe limits until we receive an advertised window */
477 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
478 tsk->rcv_win = tsk->snd_win;
480 if (tipc_sk_type_connectionless(sk)) {
481 tsk_set_unreturnable(tsk, true);
482 if (sock->type == SOCK_DGRAM)
483 tsk_set_unreliable(tsk, true);
489 static void tipc_sk_callback(struct rcu_head *head)
491 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
496 /* Caller should hold socket lock for the socket. */
497 static void __tipc_shutdown(struct socket *sock, int error)
499 struct sock *sk = sock->sk;
500 struct tipc_sock *tsk = tipc_sk(sk);
501 struct net *net = sock_net(sk);
502 long timeout = CONN_TIMEOUT_DEFAULT;
503 u32 dnode = tsk_peer_node(tsk);
506 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
507 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
508 !tsk_conn_cong(tsk)));
510 /* Reject all unreceived messages, except on an active connection
511 * (which disconnects locally & sends a 'FIN+' to peer).
513 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
514 if (TIPC_SKB_CB(skb)->bytes_read) {
518 if (!tipc_sk_type_connectionless(sk) &&
519 sk->sk_state != TIPC_DISCONNECTING) {
520 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
521 tipc_node_remove_conn(net, dnode, tsk->portid);
523 tipc_sk_respond(sk, skb, error);
526 if (tipc_sk_type_connectionless(sk))
529 if (sk->sk_state != TIPC_DISCONNECTING) {
530 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
531 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
532 tsk_own_node(tsk), tsk_peer_port(tsk),
535 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
536 tipc_node_remove_conn(net, dnode, tsk->portid);
537 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
542 * tipc_release - destroy a TIPC socket
543 * @sock: socket to destroy
545 * This routine cleans up any messages that are still queued on the socket.
546 * For DGRAM and RDM socket types, all queued messages are rejected.
547 * For SEQPACKET and STREAM socket types, the first message is rejected
548 * and any others are discarded. (If the first message on a STREAM socket
549 * is partially-read, it is discarded and the next one is rejected instead.)
551 * NOTE: Rejected messages are not necessarily returned to the sender! They
552 * are returned or discarded according to the "destination droppable" setting
553 * specified for the message by the sender.
555 * Returns 0 on success, errno otherwise
557 static int tipc_release(struct socket *sock)
559 struct sock *sk = sock->sk;
560 struct tipc_sock *tsk;
563 * Exit if socket isn't fully initialized (occurs when a failed accept()
564 * releases a pre-allocated child socket that was never used)
572 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
573 sk->sk_shutdown = SHUTDOWN_MASK;
575 tipc_sk_withdraw(tsk, 0, NULL);
576 sk_stop_timer(sk, &sk->sk_timer);
580 /* Reject any messages that accumulated in backlog queue */
582 tipc_dest_list_purge(&tsk->cong_links);
583 tsk->cong_link_cnt = 0;
584 call_rcu(&tsk->rcu, tipc_sk_callback);
591 * tipc_bind - associate or disassocate TIPC name(s) with a socket
592 * @sock: socket structure
593 * @uaddr: socket address describing name(s) and desired operation
594 * @uaddr_len: size of socket address data structure
596 * Name and name sequence binding is indicated using a positive scope value;
597 * a negative scope value unbinds the specified name. Specifying no name
598 * (i.e. a socket address length of 0) unbinds all names from the socket.
600 * Returns 0 on success, errno otherwise
602 * NOTE: This routine doesn't need to take the socket lock since it doesn't
603 * access any non-constant socket information.
605 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
608 struct sock *sk = sock->sk;
609 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
610 struct tipc_sock *tsk = tipc_sk(sk);
614 if (unlikely(!uaddr_len)) {
615 res = tipc_sk_withdraw(tsk, 0, NULL);
622 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
626 if (addr->family != AF_TIPC) {
631 if (addr->addrtype == TIPC_ADDR_NAME)
632 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
633 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
638 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
639 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
640 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
645 res = (addr->scope >= 0) ?
646 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
647 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
654 * tipc_getname - get port ID of socket or peer socket
655 * @sock: socket structure
656 * @uaddr: area for returned socket address
657 * @uaddr_len: area for returned length of socket address
658 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
660 * Returns 0 on success, errno otherwise
662 * NOTE: This routine doesn't need to take the socket lock since it only
663 * accesses socket information that is unchanging (or which changes in
664 * a completely predictable manner).
666 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
669 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
670 struct sock *sk = sock->sk;
671 struct tipc_sock *tsk = tipc_sk(sk);
673 memset(addr, 0, sizeof(*addr));
675 if ((!tipc_sk_connected(sk)) &&
676 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
678 addr->addr.id.ref = tsk_peer_port(tsk);
679 addr->addr.id.node = tsk_peer_node(tsk);
681 addr->addr.id.ref = tsk->portid;
682 addr->addr.id.node = tipc_own_addr(sock_net(sk));
685 addr->addrtype = TIPC_ADDR_ID;
686 addr->family = AF_TIPC;
688 addr->addr.name.domain = 0;
690 return sizeof(*addr);
694 * tipc_poll - read and possibly block on pollmask
695 * @file: file structure associated with the socket
696 * @sock: socket for which to calculate the poll bits
699 * Returns pollmask value
702 * It appears that the usual socket locking mechanisms are not useful here
703 * since the pollmask info is potentially out-of-date the moment this routine
704 * exits. TCP and other protocols seem to rely on higher level poll routines
705 * to handle any preventable race conditions, so TIPC will do the same ...
707 * IMPORTANT: The fact that a read or write operation is indicated does NOT
708 * imply that the operation will succeed, merely that it should be performed
709 * and will not block.
711 static __poll_t tipc_poll(struct file *file, struct socket *sock,
714 struct sock *sk = sock->sk;
715 struct tipc_sock *tsk = tipc_sk(sk);
716 __poll_t revents = 0;
718 sock_poll_wait(file, sock, wait);
720 if (sk->sk_shutdown & RCV_SHUTDOWN)
721 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
722 if (sk->sk_shutdown == SHUTDOWN_MASK)
725 switch (sk->sk_state) {
726 case TIPC_ESTABLISHED:
727 case TIPC_CONNECTING:
728 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
732 if (!skb_queue_empty(&sk->sk_receive_queue))
733 revents |= EPOLLIN | EPOLLRDNORM;
736 if (tsk->group_is_open && !tsk->cong_link_cnt)
738 if (!tipc_sk_type_connectionless(sk))
740 if (skb_queue_empty(&sk->sk_receive_queue))
742 revents |= EPOLLIN | EPOLLRDNORM;
744 case TIPC_DISCONNECTING:
745 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
752 * tipc_sendmcast - send multicast message
753 * @sock: socket structure
754 * @seq: destination address
755 * @msg: message to send
756 * @dlen: length of data to send
757 * @timeout: timeout to wait for wakeup
759 * Called from function tipc_sendmsg(), which has done all sanity checks
760 * Returns the number of bytes sent on success, or errno
762 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
763 struct msghdr *msg, size_t dlen, long timeout)
765 struct sock *sk = sock->sk;
766 struct tipc_sock *tsk = tipc_sk(sk);
767 struct tipc_msg *hdr = &tsk->phdr;
768 struct net *net = sock_net(sk);
769 int mtu = tipc_bcast_get_mtu(net);
770 struct tipc_mc_method *method = &tsk->mc_method;
771 struct sk_buff_head pkts;
772 struct tipc_nlist dsts;
778 /* Block or return if any destination link is congested */
779 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
783 /* Lookup destination nodes */
784 tipc_nlist_init(&dsts, tipc_own_addr(net));
785 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
787 if (!dsts.local && !dsts.remote)
788 return -EHOSTUNREACH;
790 /* Build message header */
791 msg_set_type(hdr, TIPC_MCAST_MSG);
792 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
793 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
794 msg_set_destport(hdr, 0);
795 msg_set_destnode(hdr, 0);
796 msg_set_nametype(hdr, seq->type);
797 msg_set_namelower(hdr, seq->lower);
798 msg_set_nameupper(hdr, seq->upper);
800 /* Build message as chain of buffers */
801 skb_queue_head_init(&pkts);
802 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
804 /* Send message if build was successful */
805 if (unlikely(rc == dlen))
806 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
807 &tsk->cong_link_cnt);
809 tipc_nlist_purge(&dsts);
811 return rc ? rc : dlen;
815 * tipc_send_group_msg - send a message to a member in the group
816 * @net: network namespace
817 * @m: message to send
819 * @dnode: destination node
820 * @dport: destination port
821 * @dlen: total length of message data
823 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
824 struct msghdr *m, struct tipc_member *mb,
825 u32 dnode, u32 dport, int dlen)
827 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
828 struct tipc_mc_method *method = &tsk->mc_method;
829 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
830 struct tipc_msg *hdr = &tsk->phdr;
831 struct sk_buff_head pkts;
834 /* Complete message header */
835 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
836 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
837 msg_set_destport(hdr, dport);
838 msg_set_destnode(hdr, dnode);
839 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
841 /* Build message as chain of buffers */
842 skb_queue_head_init(&pkts);
843 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
844 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
845 if (unlikely(rc != dlen))
849 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
850 if (unlikely(rc == -ELINKCONG)) {
851 tipc_dest_push(&tsk->cong_links, dnode, 0);
852 tsk->cong_link_cnt++;
855 /* Update send window */
856 tipc_group_update_member(mb, blks);
858 /* A broadcast sent within next EXPIRE period must follow same path */
859 method->rcast = true;
860 method->mandatory = true;
865 * tipc_send_group_unicast - send message to a member in the group
866 * @sock: socket structure
867 * @m: message to send
868 * @dlen: total length of message data
869 * @timeout: timeout to wait for wakeup
871 * Called from function tipc_sendmsg(), which has done all sanity checks
872 * Returns the number of bytes sent on success, or errno
874 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
875 int dlen, long timeout)
877 struct sock *sk = sock->sk;
878 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
879 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
880 struct tipc_sock *tsk = tipc_sk(sk);
881 struct net *net = sock_net(sk);
882 struct tipc_member *mb = NULL;
886 node = dest->addr.id.node;
887 port = dest->addr.id.ref;
889 return -EHOSTUNREACH;
891 /* Block or return if destination link or member is congested */
892 rc = tipc_wait_for_cond(sock, &timeout,
893 !tipc_dest_find(&tsk->cong_links, node, 0) &&
895 !tipc_group_cong(tsk->group, node, port, blks,
901 return -EHOSTUNREACH;
903 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
905 return rc ? rc : dlen;
909 * tipc_send_group_anycast - send message to any member with given identity
910 * @sock: socket structure
911 * @m: message to send
912 * @dlen: total length of message data
913 * @timeout: timeout to wait for wakeup
915 * Called from function tipc_sendmsg(), which has done all sanity checks
916 * Returns the number of bytes sent on success, or errno
918 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
919 int dlen, long timeout)
921 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
922 struct sock *sk = sock->sk;
923 struct tipc_sock *tsk = tipc_sk(sk);
924 struct list_head *cong_links = &tsk->cong_links;
925 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
926 struct tipc_msg *hdr = &tsk->phdr;
927 struct tipc_member *first = NULL;
928 struct tipc_member *mbr = NULL;
929 struct net *net = sock_net(sk);
930 u32 node, port, exclude;
931 struct list_head dsts;
932 u32 type, inst, scope;
937 INIT_LIST_HEAD(&dsts);
939 type = msg_nametype(hdr);
940 inst = dest->addr.name.name.instance;
941 scope = msg_lookup_scope(hdr);
943 while (++lookups < 4) {
944 exclude = tipc_group_exclude(tsk->group);
948 /* Look for a non-congested destination member, if any */
950 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
951 &dstcnt, exclude, false))
952 return -EHOSTUNREACH;
953 tipc_dest_pop(&dsts, &node, &port);
954 cong = tipc_group_cong(tsk->group, node, port, blks,
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) &&
975 !tipc_group_cong(tsk->group, node, port,
980 /* Send, unless destination disappeared while waiting */
985 if (unlikely(lookups >= 4))
986 return -EHOSTUNREACH;
988 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
990 return rc ? rc : dlen;
994 * tipc_send_group_bcast - send message to all members in communication group
995 * @sk: socket structure
996 * @m: message to send
997 * @dlen: total length of message data
998 * @timeout: timeout to wait for wakeup
1000 * Called from function tipc_sendmsg(), which has done all sanity checks
1001 * Returns the number of bytes sent on success, or errno
1003 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1004 int dlen, long timeout)
1006 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1007 struct sock *sk = sock->sk;
1008 struct net *net = sock_net(sk);
1009 struct tipc_sock *tsk = tipc_sk(sk);
1010 struct tipc_nlist *dsts = tipc_group_dests(tsk->group);
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,
1024 !tsk->cong_link_cnt && tsk->group &&
1025 !tipc_group_bc_cong(tsk->group, blks));
1029 /* Complete message header */
1031 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1032 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1034 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1035 msg_set_nameinst(hdr, 0);
1037 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1038 msg_set_destport(hdr, 0);
1039 msg_set_destnode(hdr, 0);
1040 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1042 /* Avoid getting stuck with repeated forced replicasts */
1043 msg_set_grp_bc_ack_req(hdr, ack);
1045 /* Build message as chain of buffers */
1046 skb_queue_head_init(&pkts);
1047 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1048 if (unlikely(rc != dlen))
1052 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1056 /* Update broadcast sequence number and send windows */
1057 tipc_group_update_bc_members(tsk->group, blks, ack);
1059 /* Broadcast link is now free to choose method for next broadcast */
1060 method->mandatory = false;
1061 method->expires = jiffies;
1067 * tipc_send_group_mcast - send message to all members with given identity
1068 * @sock: socket structure
1069 * @m: message to send
1070 * @dlen: total length of message data
1071 * @timeout: timeout to wait for wakeup
1073 * Called from function tipc_sendmsg(), which has done all sanity checks
1074 * Returns the number of bytes sent on success, or errno
1076 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1077 int dlen, long timeout)
1079 struct sock *sk = sock->sk;
1080 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1081 struct tipc_sock *tsk = tipc_sk(sk);
1082 struct tipc_group *grp = tsk->group;
1083 struct tipc_msg *hdr = &tsk->phdr;
1084 struct net *net = sock_net(sk);
1085 u32 type, inst, scope, exclude;
1086 struct list_head dsts;
1089 INIT_LIST_HEAD(&dsts);
1091 type = msg_nametype(hdr);
1092 inst = dest->addr.name.name.instance;
1093 scope = msg_lookup_scope(hdr);
1094 exclude = tipc_group_exclude(grp);
1096 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1097 &dstcnt, exclude, true))
1098 return -EHOSTUNREACH;
1101 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1102 return tipc_send_group_unicast(sock, m, dlen, timeout);
1105 tipc_dest_list_purge(&dsts);
1106 return tipc_send_group_bcast(sock, m, dlen, timeout);
1110 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1111 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1112 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1114 * Multi-threaded: parallel calls with reference to same queues may occur
1116 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1117 struct sk_buff_head *inputq)
1119 u32 self = tipc_own_addr(net);
1120 u32 type, lower, upper, scope;
1121 struct sk_buff *skb, *_skb;
1123 struct sk_buff_head tmpq;
1124 struct list_head dports;
1125 struct tipc_msg *hdr;
1126 int user, mtyp, hlen;
1129 __skb_queue_head_init(&tmpq);
1130 INIT_LIST_HEAD(&dports);
1132 skb = tipc_skb_peek(arrvq, &inputq->lock);
1133 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1135 user = msg_user(hdr);
1136 mtyp = msg_type(hdr);
1137 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1138 onode = msg_orignode(hdr);
1139 type = msg_nametype(hdr);
1141 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1142 spin_lock_bh(&inputq->lock);
1143 if (skb_peek(arrvq) == skb) {
1144 __skb_dequeue(arrvq);
1145 __skb_queue_tail(inputq, skb);
1148 spin_unlock_bh(&inputq->lock);
1152 /* Group messages require exact scope match */
1153 if (msg_in_group(hdr)) {
1156 scope = msg_lookup_scope(hdr);
1159 /* TIPC_NODE_SCOPE means "any scope" in this context */
1161 scope = TIPC_NODE_SCOPE;
1163 scope = TIPC_CLUSTER_SCOPE;
1165 lower = msg_namelower(hdr);
1166 upper = msg_nameupper(hdr);
1169 /* Create destination port list: */
1170 tipc_nametbl_mc_lookup(net, type, lower, upper,
1171 scope, exact, &dports);
1173 /* Clone message per destination */
1174 while (tipc_dest_pop(&dports, NULL, &portid)) {
1175 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1177 msg_set_destport(buf_msg(_skb), portid);
1178 __skb_queue_tail(&tmpq, _skb);
1181 pr_warn("Failed to clone mcast rcv buffer\n");
1183 /* Append to inputq if not already done by other thread */
1184 spin_lock_bh(&inputq->lock);
1185 if (skb_peek(arrvq) == skb) {
1186 skb_queue_splice_tail_init(&tmpq, inputq);
1187 kfree_skb(__skb_dequeue(arrvq));
1189 spin_unlock_bh(&inputq->lock);
1190 __skb_queue_purge(&tmpq);
1193 tipc_sk_rcv(net, inputq);
1197 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1198 * @tsk: receiving socket
1199 * @skb: pointer to message buffer.
1201 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1202 struct sk_buff_head *inputq,
1203 struct sk_buff_head *xmitq)
1205 struct tipc_msg *hdr = buf_msg(skb);
1206 u32 onode = tsk_own_node(tsk);
1207 struct sock *sk = &tsk->sk;
1208 int mtyp = msg_type(hdr);
1211 /* Ignore if connection cannot be validated: */
1212 if (!tsk_peer_msg(tsk, hdr))
1215 if (unlikely(msg_errcode(hdr))) {
1216 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1217 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1218 tsk_peer_port(tsk));
1219 sk->sk_state_change(sk);
1221 /* State change is ignored if socket already awake,
1222 * - convert msg to abort msg and add to inqueue
1224 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1225 msg_set_type(hdr, TIPC_CONN_MSG);
1226 msg_set_size(hdr, BASIC_H_SIZE);
1227 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1228 __skb_queue_tail(inputq, skb);
1232 tsk->probe_unacked = false;
1234 if (mtyp == CONN_PROBE) {
1235 msg_set_type(hdr, CONN_PROBE_REPLY);
1236 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1237 __skb_queue_tail(xmitq, skb);
1239 } else if (mtyp == CONN_ACK) {
1240 conn_cong = tsk_conn_cong(tsk);
1241 tsk->snt_unacked -= msg_conn_ack(hdr);
1242 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1243 tsk->snd_win = msg_adv_win(hdr);
1245 sk->sk_write_space(sk);
1246 } else if (mtyp != CONN_PROBE_REPLY) {
1247 pr_warn("Received unknown CONN_PROTO msg\n");
1254 * tipc_sendmsg - send message in connectionless manner
1255 * @sock: socket structure
1256 * @m: message to send
1257 * @dsz: amount of user data to be sent
1259 * Message must have an destination specified explicitly.
1260 * Used for SOCK_RDM and SOCK_DGRAM messages,
1261 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1262 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1264 * Returns the number of bytes sent on success, or errno otherwise
1266 static int tipc_sendmsg(struct socket *sock,
1267 struct msghdr *m, size_t dsz)
1269 struct sock *sk = sock->sk;
1273 ret = __tipc_sendmsg(sock, m, dsz);
1279 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1281 struct sock *sk = sock->sk;
1282 struct net *net = sock_net(sk);
1283 struct tipc_sock *tsk = tipc_sk(sk);
1284 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1285 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1286 struct list_head *clinks = &tsk->cong_links;
1287 bool syn = !tipc_sk_type_connectionless(sk);
1288 struct tipc_group *grp = tsk->group;
1289 struct tipc_msg *hdr = &tsk->phdr;
1290 struct tipc_name_seq *seq;
1291 struct sk_buff_head pkts;
1292 u32 dport, dnode = 0;
1296 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1300 if (unlikely(m->msg_namelen < sizeof(*dest)))
1302 if (unlikely(dest->family != AF_TIPC))
1308 return tipc_send_group_bcast(sock, m, dlen, timeout);
1309 if (dest->addrtype == TIPC_ADDR_NAME)
1310 return tipc_send_group_anycast(sock, m, dlen, timeout);
1311 if (dest->addrtype == TIPC_ADDR_ID)
1312 return tipc_send_group_unicast(sock, m, dlen, timeout);
1313 if (dest->addrtype == TIPC_ADDR_MCAST)
1314 return tipc_send_group_mcast(sock, m, dlen, timeout);
1318 if (unlikely(!dest)) {
1320 if (!syn || dest->family != AF_TIPC)
1321 return -EDESTADDRREQ;
1324 if (unlikely(syn)) {
1325 if (sk->sk_state == TIPC_LISTEN)
1327 if (sk->sk_state != TIPC_OPEN)
1331 if (dest->addrtype == TIPC_ADDR_NAME) {
1332 tsk->conn_type = dest->addr.name.name.type;
1333 tsk->conn_instance = dest->addr.name.name.instance;
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))
1378 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1379 if (unlikely(rc == -ELINKCONG)) {
1380 tipc_dest_push(clinks, dnode, 0);
1381 tsk->cong_link_cnt++;
1385 if (unlikely(syn && !rc))
1386 tipc_set_sk_state(sk, TIPC_CONNECTING);
1388 return rc ? rc : dlen;
1392 * tipc_sendstream - send stream-oriented data
1393 * @sock: socket structure
1395 * @dsz: total length of data to be transmitted
1397 * Used for SOCK_STREAM data.
1399 * Returns the number of bytes sent on success (or partial success),
1400 * or errno if no data sent
1402 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1404 struct sock *sk = sock->sk;
1408 ret = __tipc_sendstream(sock, m, dsz);
1414 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1416 struct sock *sk = sock->sk;
1417 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1418 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1419 struct tipc_sock *tsk = tipc_sk(sk);
1420 struct tipc_msg *hdr = &tsk->phdr;
1421 struct net *net = sock_net(sk);
1422 struct sk_buff_head pkts;
1423 u32 dnode = tsk_peer_node(tsk);
1427 skb_queue_head_init(&pkts);
1429 if (unlikely(dlen > INT_MAX))
1432 /* Handle implicit connection setup */
1433 if (unlikely(dest)) {
1434 rc = __tipc_sendmsg(sock, m, dlen);
1435 if (dlen && dlen == rc) {
1436 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1437 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1443 rc = tipc_wait_for_cond(sock, &timeout,
1444 (!tsk->cong_link_cnt &&
1445 !tsk_conn_cong(tsk) &&
1446 tipc_sk_connected(sk)));
1450 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1451 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1452 if (unlikely(rc != send))
1455 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1456 if (unlikely(rc == -ELINKCONG)) {
1457 tsk->cong_link_cnt = 1;
1461 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1464 } while (sent < dlen && !rc);
1466 return sent ? sent : rc;
1470 * tipc_send_packet - send a connection-oriented message
1471 * @sock: socket structure
1472 * @m: message to send
1473 * @dsz: length of data to be transmitted
1475 * Used for SOCK_SEQPACKET messages.
1477 * Returns the number of bytes sent on success, or errno otherwise
1479 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1481 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1484 return tipc_sendstream(sock, m, dsz);
1487 /* tipc_sk_finish_conn - complete the setup of a connection
1489 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1492 struct sock *sk = &tsk->sk;
1493 struct net *net = sock_net(sk);
1494 struct tipc_msg *msg = &tsk->phdr;
1496 msg_set_destnode(msg, peer_node);
1497 msg_set_destport(msg, peer_port);
1498 msg_set_type(msg, TIPC_CONN_MSG);
1499 msg_set_lookup_scope(msg, 0);
1500 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1502 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1503 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1504 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1505 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1506 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1507 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1510 /* Fall back to message based flow control */
1511 tsk->rcv_win = FLOWCTL_MSG_WIN;
1512 tsk->snd_win = FLOWCTL_MSG_WIN;
1516 * tipc_sk_set_orig_addr - capture sender's address for received message
1517 * @m: descriptor for message info
1518 * @hdr: received message header
1520 * Note: Address is not captured if not requested by receiver.
1522 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1524 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1525 struct tipc_msg *hdr = buf_msg(skb);
1530 srcaddr->sock.family = AF_TIPC;
1531 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1532 srcaddr->sock.scope = 0;
1533 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1534 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1535 srcaddr->sock.addr.name.domain = 0;
1536 m->msg_namelen = sizeof(struct sockaddr_tipc);
1538 if (!msg_in_group(hdr))
1541 /* Group message users may also want to know sending member's id */
1542 srcaddr->member.family = AF_TIPC;
1543 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1544 srcaddr->member.scope = 0;
1545 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1546 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1547 srcaddr->member.addr.name.domain = 0;
1548 m->msg_namelen = sizeof(*srcaddr);
1552 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1553 * @m: descriptor for message info
1554 * @skb: received message buffer
1555 * @tsk: TIPC port associated with message
1557 * Note: Ancillary data is not captured if not requested by receiver.
1559 * Returns 0 if successful, otherwise errno
1561 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1562 struct tipc_sock *tsk)
1564 struct tipc_msg *msg;
1571 if (likely(m->msg_controllen == 0))
1575 /* Optionally capture errored message object(s) */
1576 err = msg ? msg_errcode(msg) : 0;
1577 if (unlikely(err)) {
1579 anc_data[1] = msg_data_sz(msg);
1580 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1584 if (skb_linearize(skb))
1587 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1594 /* Optionally capture message destination object */
1595 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1596 switch (dest_type) {
1597 case TIPC_NAMED_MSG:
1599 anc_data[0] = msg_nametype(msg);
1600 anc_data[1] = msg_namelower(msg);
1601 anc_data[2] = msg_namelower(msg);
1603 case TIPC_MCAST_MSG:
1605 anc_data[0] = msg_nametype(msg);
1606 anc_data[1] = msg_namelower(msg);
1607 anc_data[2] = msg_nameupper(msg);
1610 has_name = (tsk->conn_type != 0);
1611 anc_data[0] = tsk->conn_type;
1612 anc_data[1] = tsk->conn_instance;
1613 anc_data[2] = tsk->conn_instance;
1619 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1627 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1629 struct sock *sk = &tsk->sk;
1630 struct net *net = sock_net(sk);
1631 struct sk_buff *skb = NULL;
1632 struct tipc_msg *msg;
1633 u32 peer_port = tsk_peer_port(tsk);
1634 u32 dnode = tsk_peer_node(tsk);
1636 if (!tipc_sk_connected(sk))
1638 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1639 dnode, tsk_own_node(tsk), peer_port,
1640 tsk->portid, TIPC_OK);
1644 msg_set_conn_ack(msg, tsk->rcv_unacked);
1645 tsk->rcv_unacked = 0;
1647 /* Adjust to and advertize the correct window limit */
1648 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1649 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1650 msg_set_adv_win(msg, tsk->rcv_win);
1652 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1655 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1657 struct sock *sk = sock->sk;
1659 long timeo = *timeop;
1660 int err = sock_error(sk);
1666 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1667 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1668 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1673 timeo = schedule_timeout(timeo);
1677 if (!skb_queue_empty(&sk->sk_receive_queue))
1682 err = sock_intr_errno(timeo);
1683 if (signal_pending(current))
1686 err = sock_error(sk);
1690 finish_wait(sk_sleep(sk), &wait);
1696 * tipc_recvmsg - receive packet-oriented message
1697 * @m: descriptor for message info
1698 * @buflen: length of user buffer area
1699 * @flags: receive flags
1701 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1702 * If the complete message doesn't fit in user area, truncate it.
1704 * Returns size of returned message data, errno otherwise
1706 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1707 size_t buflen, int flags)
1709 struct sock *sk = sock->sk;
1710 bool connected = !tipc_sk_type_connectionless(sk);
1711 struct tipc_sock *tsk = tipc_sk(sk);
1712 int rc, err, hlen, dlen, copy;
1713 struct sk_buff_head xmitq;
1714 struct tipc_msg *hdr;
1715 struct sk_buff *skb;
1719 /* Catch invalid receive requests */
1720 if (unlikely(!buflen))
1724 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1728 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1730 /* Step rcv queue to first msg with data or error; wait if necessary */
1732 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1735 skb = skb_peek(&sk->sk_receive_queue);
1737 dlen = msg_data_sz(hdr);
1738 hlen = msg_hdr_sz(hdr);
1739 err = msg_errcode(hdr);
1740 grp_evt = msg_is_grp_evt(hdr);
1741 if (likely(dlen || err))
1743 tsk_advance_rx_queue(sk);
1746 /* Collect msg meta data, including error code and rejected data */
1747 tipc_sk_set_orig_addr(m, skb);
1748 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1753 /* Capture data if non-error msg, otherwise just set return value */
1755 copy = min_t(int, dlen, buflen);
1756 if (unlikely(copy != dlen))
1757 m->msg_flags |= MSG_TRUNC;
1758 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1762 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1768 /* Mark message as group event if applicable */
1769 if (unlikely(grp_evt)) {
1770 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1771 m->msg_flags |= MSG_EOR;
1772 m->msg_flags |= MSG_OOB;
1776 /* Caption of data or error code/rejected data was successful */
1777 if (unlikely(flags & MSG_PEEK))
1780 /* Send group flow control advertisement when applicable */
1781 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1782 skb_queue_head_init(&xmitq);
1783 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1784 msg_orignode(hdr), msg_origport(hdr),
1786 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1789 tsk_advance_rx_queue(sk);
1791 if (likely(!connected))
1794 /* Send connection flow control advertisement when applicable */
1795 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1796 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1797 tipc_sk_send_ack(tsk);
1800 return rc ? rc : copy;
1804 * tipc_recvstream - receive stream-oriented data
1805 * @m: descriptor for message info
1806 * @buflen: total size of user buffer area
1807 * @flags: receive flags
1809 * Used for SOCK_STREAM messages only. If not enough data is available
1810 * will optionally wait for more; never truncates data.
1812 * Returns size of returned message data, errno otherwise
1814 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1815 size_t buflen, int flags)
1817 struct sock *sk = sock->sk;
1818 struct tipc_sock *tsk = tipc_sk(sk);
1819 struct sk_buff *skb;
1820 struct tipc_msg *hdr;
1821 struct tipc_skb_cb *skb_cb;
1822 bool peek = flags & MSG_PEEK;
1823 int offset, required, copy, copied = 0;
1824 int hlen, dlen, err, rc;
1827 /* Catch invalid receive attempts */
1828 if (unlikely(!buflen))
1833 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1837 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1838 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1841 /* Look at first msg in receive queue; wait if necessary */
1842 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1845 skb = skb_peek(&sk->sk_receive_queue);
1846 skb_cb = TIPC_SKB_CB(skb);
1848 dlen = msg_data_sz(hdr);
1849 hlen = msg_hdr_sz(hdr);
1850 err = msg_errcode(hdr);
1852 /* Discard any empty non-errored (SYN-) message */
1853 if (unlikely(!dlen && !err)) {
1854 tsk_advance_rx_queue(sk);
1858 /* Collect msg meta data, incl. error code and rejected data */
1860 tipc_sk_set_orig_addr(m, skb);
1861 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1867 /* Copy data if msg ok, otherwise return error/partial data */
1869 offset = skb_cb->bytes_read;
1870 copy = min_t(int, dlen - offset, buflen - copied);
1871 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1876 if (unlikely(offset < dlen)) {
1878 skb_cb->bytes_read = offset;
1883 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1892 tsk_advance_rx_queue(sk);
1894 /* Send connection flow control advertisement when applicable */
1895 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1896 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1897 tipc_sk_send_ack(tsk);
1899 /* Exit if all requested data or FIN/error received */
1900 if (copied == buflen || err)
1903 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1906 return copied ? copied : rc;
1910 * tipc_write_space - wake up thread if port congestion is released
1913 static void tipc_write_space(struct sock *sk)
1915 struct socket_wq *wq;
1918 wq = rcu_dereference(sk->sk_wq);
1919 if (skwq_has_sleeper(wq))
1920 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1921 EPOLLWRNORM | EPOLLWRBAND);
1926 * tipc_data_ready - wake up threads to indicate messages have been received
1928 * @len: the length of messages
1930 static void tipc_data_ready(struct sock *sk)
1932 struct socket_wq *wq;
1935 wq = rcu_dereference(sk->sk_wq);
1936 if (skwq_has_sleeper(wq))
1937 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1938 EPOLLRDNORM | EPOLLRDBAND);
1942 static void tipc_sock_destruct(struct sock *sk)
1944 __skb_queue_purge(&sk->sk_receive_queue);
1947 static void tipc_sk_proto_rcv(struct sock *sk,
1948 struct sk_buff_head *inputq,
1949 struct sk_buff_head *xmitq)
1951 struct sk_buff *skb = __skb_dequeue(inputq);
1952 struct tipc_sock *tsk = tipc_sk(sk);
1953 struct tipc_msg *hdr = buf_msg(skb);
1954 struct tipc_group *grp = tsk->group;
1955 bool wakeup = false;
1957 switch (msg_user(hdr)) {
1959 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1962 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1963 tsk->cong_link_cnt--;
1966 case GROUP_PROTOCOL:
1967 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1970 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1971 hdr, inputq, xmitq);
1978 sk->sk_write_space(sk);
1984 * tipc_filter_connect - Handle incoming message for a connection-based socket
1986 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1988 * Returns true if everything ok, false otherwise
1990 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1992 struct sock *sk = &tsk->sk;
1993 struct net *net = sock_net(sk);
1994 struct tipc_msg *hdr = buf_msg(skb);
1995 u32 pport = msg_origport(hdr);
1996 u32 pnode = msg_orignode(hdr);
1998 if (unlikely(msg_mcast(hdr)))
2001 switch (sk->sk_state) {
2002 case TIPC_CONNECTING:
2003 /* Accept only ACK or NACK message */
2004 if (unlikely(!msg_connected(hdr))) {
2005 if (pport != tsk_peer_port(tsk) ||
2006 pnode != tsk_peer_node(tsk))
2009 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2010 sk->sk_err = ECONNREFUSED;
2011 sk->sk_state_change(sk);
2015 if (unlikely(msg_errcode(hdr))) {
2016 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2017 sk->sk_err = ECONNREFUSED;
2018 sk->sk_state_change(sk);
2022 if (unlikely(!msg_isdata(hdr))) {
2023 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2024 sk->sk_err = EINVAL;
2025 sk->sk_state_change(sk);
2029 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2030 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2032 /* If 'ACK+' message, add to socket receive queue */
2033 if (msg_data_sz(hdr))
2036 /* If empty 'ACK-' message, wake up sleeping connect() */
2037 sk->sk_data_ready(sk);
2039 /* 'ACK-' message is neither accepted nor rejected: */
2040 msg_set_dest_droppable(hdr, 1);
2044 case TIPC_DISCONNECTING:
2047 /* Accept only SYN message */
2048 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2051 case TIPC_ESTABLISHED:
2052 /* Accept only connection-based messages sent by peer */
2053 if (unlikely(!tsk_peer_msg(tsk, hdr)))
2056 if (unlikely(msg_errcode(hdr))) {
2057 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2058 /* Let timer expire on it's own */
2059 tipc_node_remove_conn(net, tsk_peer_node(tsk),
2061 sk->sk_state_change(sk);
2065 pr_err("Unknown sk_state %u\n", sk->sk_state);
2072 * rcvbuf_limit - get proper overload limit of socket receive queue
2076 * For connection oriented messages, irrespective of importance,
2077 * default queue limit is 2 MB.
2079 * For connectionless messages, queue limits are based on message
2080 * importance as follows:
2082 * TIPC_LOW_IMPORTANCE (2 MB)
2083 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2084 * TIPC_HIGH_IMPORTANCE (8 MB)
2085 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2087 * Returns overload limit according to corresponding message importance
2089 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2091 struct tipc_sock *tsk = tipc_sk(sk);
2092 struct tipc_msg *hdr = buf_msg(skb);
2094 if (unlikely(msg_in_group(hdr)))
2095 return sk->sk_rcvbuf;
2097 if (unlikely(!msg_connected(hdr)))
2098 return sk->sk_rcvbuf << msg_importance(hdr);
2100 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2101 return sk->sk_rcvbuf;
2103 return FLOWCTL_MSG_LIM;
2107 * tipc_sk_filter_rcv - validate incoming message
2109 * @skb: pointer to message.
2111 * Enqueues message on receive queue if acceptable; optionally handles
2112 * disconnect indication for a connected socket.
2114 * Called with socket lock already taken
2117 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2118 struct sk_buff_head *xmitq)
2120 bool sk_conn = !tipc_sk_type_connectionless(sk);
2121 struct tipc_sock *tsk = tipc_sk(sk);
2122 struct tipc_group *grp = tsk->group;
2123 struct tipc_msg *hdr = buf_msg(skb);
2124 struct net *net = sock_net(sk);
2125 struct sk_buff_head inputq;
2126 int limit, err = TIPC_OK;
2128 TIPC_SKB_CB(skb)->bytes_read = 0;
2129 __skb_queue_head_init(&inputq);
2130 __skb_queue_tail(&inputq, skb);
2132 if (unlikely(!msg_isdata(hdr)))
2133 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2136 tipc_group_filter_msg(grp, &inputq, xmitq);
2138 /* Validate and add to receive buffer if there is space */
2139 while ((skb = __skb_dequeue(&inputq))) {
2141 limit = rcvbuf_limit(sk, skb);
2142 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2143 (!sk_conn && msg_connected(hdr)) ||
2144 (!grp && msg_in_group(hdr)))
2145 err = TIPC_ERR_NO_PORT;
2146 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2147 atomic_inc(&sk->sk_drops);
2148 err = TIPC_ERR_OVERLOAD;
2151 if (unlikely(err)) {
2152 tipc_skb_reject(net, err, skb, xmitq);
2156 __skb_queue_tail(&sk->sk_receive_queue, skb);
2157 skb_set_owner_r(skb, sk);
2158 sk->sk_data_ready(sk);
2163 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2167 * Caller must hold socket lock
2169 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2171 unsigned int before = sk_rmem_alloc_get(sk);
2172 struct sk_buff_head xmitq;
2175 __skb_queue_head_init(&xmitq);
2177 tipc_sk_filter_rcv(sk, skb, &xmitq);
2178 added = sk_rmem_alloc_get(sk) - before;
2179 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2181 /* Send pending response/rejected messages, if any */
2182 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2187 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2188 * inputq and try adding them to socket or backlog queue
2189 * @inputq: list of incoming buffers with potentially different destinations
2190 * @sk: socket where the buffers should be enqueued
2191 * @dport: port number for the socket
2193 * Caller must hold socket lock
2195 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2196 u32 dport, struct sk_buff_head *xmitq)
2198 unsigned long time_limit = jiffies + 2;
2199 struct sk_buff *skb;
2204 while (skb_queue_len(inputq)) {
2205 if (unlikely(time_after_eq(jiffies, time_limit)))
2208 skb = tipc_skb_dequeue(inputq, dport);
2212 /* Add message directly to receive queue if possible */
2213 if (!sock_owned_by_user(sk)) {
2214 tipc_sk_filter_rcv(sk, skb, xmitq);
2218 /* Try backlog, compensating for double-counted bytes */
2219 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2220 if (!sk->sk_backlog.len)
2221 atomic_set(dcnt, 0);
2222 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2223 if (likely(!sk_add_backlog(sk, skb, lim)))
2226 /* Overload => reject message back to sender */
2227 onode = tipc_own_addr(sock_net(sk));
2228 atomic_inc(&sk->sk_drops);
2229 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2230 __skb_queue_tail(xmitq, skb);
2236 * tipc_sk_rcv - handle a chain of incoming buffers
2237 * @inputq: buffer list containing the buffers
2238 * Consumes all buffers in list until inputq is empty
2239 * Note: may be called in multiple threads referring to the same queue
2241 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2243 struct sk_buff_head xmitq;
2244 u32 dnode, dport = 0;
2246 struct tipc_sock *tsk;
2248 struct sk_buff *skb;
2250 __skb_queue_head_init(&xmitq);
2251 while (skb_queue_len(inputq)) {
2252 dport = tipc_skb_peek_port(inputq, dport);
2253 tsk = tipc_sk_lookup(net, dport);
2257 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2258 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2259 spin_unlock_bh(&sk->sk_lock.slock);
2261 /* Send pending response/rejected messages, if any */
2262 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2266 /* No destination socket => dequeue skb if still there */
2267 skb = tipc_skb_dequeue(inputq, dport);
2271 /* Try secondary lookup if unresolved named message */
2272 err = TIPC_ERR_NO_PORT;
2273 if (tipc_msg_lookup_dest(net, skb, &err))
2276 /* Prepare for message rejection */
2277 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2280 dnode = msg_destnode(buf_msg(skb));
2281 tipc_node_xmit_skb(net, skb, dnode, dport);
2285 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2287 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2288 struct sock *sk = sock->sk;
2292 int err = sock_error(sk);
2297 if (signal_pending(current))
2298 return sock_intr_errno(*timeo_p);
2300 add_wait_queue(sk_sleep(sk), &wait);
2301 done = sk_wait_event(sk, timeo_p,
2302 sk->sk_state != TIPC_CONNECTING, &wait);
2303 remove_wait_queue(sk_sleep(sk), &wait);
2309 * tipc_connect - establish a connection to another TIPC port
2310 * @sock: socket structure
2311 * @dest: socket address for destination port
2312 * @destlen: size of socket address data structure
2313 * @flags: file-related flags associated with socket
2315 * Returns 0 on success, errno otherwise
2317 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2318 int destlen, int flags)
2320 struct sock *sk = sock->sk;
2321 struct tipc_sock *tsk = tipc_sk(sk);
2322 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2323 struct msghdr m = {NULL,};
2324 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2328 if (destlen != sizeof(struct sockaddr_tipc))
2338 if (dst->family == AF_UNSPEC) {
2339 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2340 if (!tipc_sk_type_connectionless(sk))
2343 } else if (dst->family != AF_TIPC) {
2346 if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2351 /* DGRAM/RDM connect(), just save the destaddr */
2352 if (tipc_sk_type_connectionless(sk)) {
2353 memcpy(&tsk->peer, dest, destlen);
2357 previous = sk->sk_state;
2359 switch (sk->sk_state) {
2361 /* Send a 'SYN-' to destination */
2363 m.msg_namelen = destlen;
2365 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2366 * indicate send_msg() is never blocked.
2369 m.msg_flags = MSG_DONTWAIT;
2371 res = __tipc_sendmsg(sock, &m, 0);
2372 if ((res < 0) && (res != -EWOULDBLOCK))
2375 /* Just entered TIPC_CONNECTING state; the only
2376 * difference is that return value in non-blocking
2377 * case is EINPROGRESS, rather than EALREADY.
2381 case TIPC_CONNECTING:
2383 if (previous == TIPC_CONNECTING)
2387 timeout = msecs_to_jiffies(timeout);
2388 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2389 res = tipc_wait_for_connect(sock, &timeout);
2391 case TIPC_ESTABLISHED:
2404 * tipc_listen - allow socket to listen for incoming connections
2405 * @sock: socket structure
2408 * Returns 0 on success, errno otherwise
2410 static int tipc_listen(struct socket *sock, int len)
2412 struct sock *sk = sock->sk;
2416 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2422 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2424 struct sock *sk = sock->sk;
2428 /* True wake-one mechanism for incoming connections: only
2429 * one process gets woken up, not the 'whole herd'.
2430 * Since we do not 'race & poll' for established sockets
2431 * anymore, the common case will execute the loop only once.
2434 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2435 TASK_INTERRUPTIBLE);
2436 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2438 timeo = schedule_timeout(timeo);
2442 if (!skb_queue_empty(&sk->sk_receive_queue))
2447 err = sock_intr_errno(timeo);
2448 if (signal_pending(current))
2451 finish_wait(sk_sleep(sk), &wait);
2456 * tipc_accept - wait for connection request
2457 * @sock: listening socket
2458 * @newsock: new socket that is to be connected
2459 * @flags: file-related flags associated with socket
2461 * Returns 0 on success, errno otherwise
2463 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2466 struct sock *new_sk, *sk = sock->sk;
2467 struct sk_buff *buf;
2468 struct tipc_sock *new_tsock;
2469 struct tipc_msg *msg;
2475 if (sk->sk_state != TIPC_LISTEN) {
2479 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2480 res = tipc_wait_for_accept(sock, timeo);
2484 buf = skb_peek(&sk->sk_receive_queue);
2486 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2489 security_sk_clone(sock->sk, new_sock->sk);
2491 new_sk = new_sock->sk;
2492 new_tsock = tipc_sk(new_sk);
2495 /* we lock on new_sk; but lockdep sees the lock on sk */
2496 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2499 * Reject any stray messages received by new socket
2500 * before the socket lock was taken (very, very unlikely)
2502 tsk_rej_rx_queue(new_sk);
2504 /* Connect new socket to it's peer */
2505 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2507 tsk_set_importance(new_tsock, msg_importance(msg));
2508 if (msg_named(msg)) {
2509 new_tsock->conn_type = msg_nametype(msg);
2510 new_tsock->conn_instance = msg_nameinst(msg);
2514 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2515 * Respond to 'SYN+' by queuing it on new socket.
2517 if (!msg_data_sz(msg)) {
2518 struct msghdr m = {NULL,};
2520 tsk_advance_rx_queue(sk);
2521 __tipc_sendstream(new_sock, &m, 0);
2523 __skb_dequeue(&sk->sk_receive_queue);
2524 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2525 skb_set_owner_r(buf, new_sk);
2527 release_sock(new_sk);
2534 * tipc_shutdown - shutdown socket connection
2535 * @sock: socket structure
2536 * @how: direction to close (must be SHUT_RDWR)
2538 * Terminates connection (if necessary), then purges socket's receive queue.
2540 * Returns 0 on success, errno otherwise
2542 static int tipc_shutdown(struct socket *sock, int how)
2544 struct sock *sk = sock->sk;
2547 if (how != SHUT_RDWR)
2552 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2553 sk->sk_shutdown = SEND_SHUTDOWN;
2555 if (sk->sk_state == TIPC_DISCONNECTING) {
2556 /* Discard any unreceived messages */
2557 __skb_queue_purge(&sk->sk_receive_queue);
2559 /* Wake up anyone sleeping in poll */
2560 sk->sk_state_change(sk);
2570 static void tipc_sk_timeout(struct timer_list *t)
2572 struct sock *sk = from_timer(sk, t, sk_timer);
2573 struct tipc_sock *tsk = tipc_sk(sk);
2574 u32 peer_port = tsk_peer_port(tsk);
2575 u32 peer_node = tsk_peer_node(tsk);
2576 u32 own_node = tsk_own_node(tsk);
2577 u32 own_port = tsk->portid;
2578 struct net *net = sock_net(sk);
2579 struct sk_buff *skb = NULL;
2582 if (!tipc_sk_connected(sk))
2585 /* Try again later if socket is busy */
2586 if (sock_owned_by_user(sk)) {
2587 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2591 if (tsk->probe_unacked) {
2592 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2593 tipc_node_remove_conn(net, peer_node, peer_port);
2594 sk->sk_state_change(sk);
2597 /* Send new probe */
2598 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2599 peer_node, own_node, peer_port, own_port,
2601 tsk->probe_unacked = true;
2602 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2606 tipc_node_xmit_skb(net, skb, peer_node, own_port);
2610 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2611 struct tipc_name_seq const *seq)
2613 struct sock *sk = &tsk->sk;
2614 struct net *net = sock_net(sk);
2615 struct publication *publ;
2618 if (scope != TIPC_NODE_SCOPE)
2619 scope = TIPC_CLUSTER_SCOPE;
2621 if (tipc_sk_connected(sk))
2623 key = tsk->portid + tsk->pub_count + 1;
2624 if (key == tsk->portid)
2627 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2628 scope, tsk->portid, key);
2629 if (unlikely(!publ))
2632 list_add(&publ->binding_sock, &tsk->publications);
2638 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2639 struct tipc_name_seq const *seq)
2641 struct net *net = sock_net(&tsk->sk);
2642 struct publication *publ;
2643 struct publication *safe;
2646 if (scope != TIPC_NODE_SCOPE)
2647 scope = TIPC_CLUSTER_SCOPE;
2649 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2651 if (publ->scope != scope)
2653 if (publ->type != seq->type)
2655 if (publ->lower != seq->lower)
2657 if (publ->upper != seq->upper)
2659 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2660 publ->upper, publ->key);
2664 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2665 publ->upper, publ->key);
2668 if (list_empty(&tsk->publications))
2673 /* tipc_sk_reinit: set non-zero address in all existing sockets
2674 * when we go from standalone to network mode.
2676 void tipc_sk_reinit(struct net *net)
2678 struct tipc_net *tn = net_generic(net, tipc_net_id);
2679 struct rhashtable_iter iter;
2680 struct tipc_sock *tsk;
2681 struct tipc_msg *msg;
2683 rhashtable_walk_enter(&tn->sk_rht, &iter);
2686 rhashtable_walk_start(&iter);
2688 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2689 spin_lock_bh(&tsk->sk.sk_lock.slock);
2691 msg_set_prevnode(msg, tipc_own_addr(net));
2692 msg_set_orignode(msg, tipc_own_addr(net));
2693 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2696 rhashtable_walk_stop(&iter);
2697 } while (tsk == ERR_PTR(-EAGAIN));
2699 rhashtable_walk_exit(&iter);
2702 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2704 struct tipc_net *tn = net_generic(net, tipc_net_id);
2705 struct tipc_sock *tsk;
2708 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2710 sock_hold(&tsk->sk);
2716 static int tipc_sk_insert(struct tipc_sock *tsk)
2718 struct sock *sk = &tsk->sk;
2719 struct net *net = sock_net(sk);
2720 struct tipc_net *tn = net_generic(net, tipc_net_id);
2721 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2722 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2724 while (remaining--) {
2726 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2727 portid = TIPC_MIN_PORT;
2728 tsk->portid = portid;
2729 sock_hold(&tsk->sk);
2730 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2739 static void tipc_sk_remove(struct tipc_sock *tsk)
2741 struct sock *sk = &tsk->sk;
2742 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2744 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2745 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2750 static const struct rhashtable_params tsk_rht_params = {
2752 .head_offset = offsetof(struct tipc_sock, node),
2753 .key_offset = offsetof(struct tipc_sock, portid),
2754 .key_len = sizeof(u32), /* portid */
2755 .max_size = 1048576,
2757 .automatic_shrinking = true,
2760 int tipc_sk_rht_init(struct net *net)
2762 struct tipc_net *tn = net_generic(net, tipc_net_id);
2764 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2767 void tipc_sk_rht_destroy(struct net *net)
2769 struct tipc_net *tn = net_generic(net, tipc_net_id);
2771 /* Wait for socket readers to complete */
2774 rhashtable_destroy(&tn->sk_rht);
2777 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2779 struct net *net = sock_net(&tsk->sk);
2780 struct tipc_group *grp = tsk->group;
2781 struct tipc_msg *hdr = &tsk->phdr;
2782 struct tipc_name_seq seq;
2785 if (mreq->type < TIPC_RESERVED_TYPES)
2787 if (mreq->scope > TIPC_NODE_SCOPE)
2791 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2795 msg_set_lookup_scope(hdr, mreq->scope);
2796 msg_set_nametype(hdr, mreq->type);
2797 msg_set_dest_droppable(hdr, true);
2798 seq.type = mreq->type;
2799 seq.lower = mreq->instance;
2800 seq.upper = seq.lower;
2801 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2802 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2804 tipc_group_delete(net, grp);
2808 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2809 tsk->mc_method.rcast = true;
2810 tsk->mc_method.mandatory = true;
2811 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2815 static int tipc_sk_leave(struct tipc_sock *tsk)
2817 struct net *net = sock_net(&tsk->sk);
2818 struct tipc_group *grp = tsk->group;
2819 struct tipc_name_seq seq;
2824 tipc_group_self(grp, &seq, &scope);
2825 tipc_group_delete(net, grp);
2827 tipc_sk_withdraw(tsk, scope, &seq);
2832 * tipc_setsockopt - set socket option
2833 * @sock: socket structure
2834 * @lvl: option level
2835 * @opt: option identifier
2836 * @ov: pointer to new option value
2837 * @ol: length of option value
2839 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2840 * (to ease compatibility).
2842 * Returns 0 on success, errno otherwise
2844 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2845 char __user *ov, unsigned int ol)
2847 struct sock *sk = sock->sk;
2848 struct tipc_sock *tsk = tipc_sk(sk);
2849 struct tipc_group_req mreq;
2853 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2855 if (lvl != SOL_TIPC)
2856 return -ENOPROTOOPT;
2859 case TIPC_IMPORTANCE:
2860 case TIPC_SRC_DROPPABLE:
2861 case TIPC_DEST_DROPPABLE:
2862 case TIPC_CONN_TIMEOUT:
2863 if (ol < sizeof(value))
2865 if (get_user(value, (u32 __user *)ov))
2868 case TIPC_GROUP_JOIN:
2869 if (ol < sizeof(mreq))
2871 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2882 case TIPC_IMPORTANCE:
2883 res = tsk_set_importance(tsk, value);
2885 case TIPC_SRC_DROPPABLE:
2886 if (sock->type != SOCK_STREAM)
2887 tsk_set_unreliable(tsk, value);
2891 case TIPC_DEST_DROPPABLE:
2892 tsk_set_unreturnable(tsk, value);
2894 case TIPC_CONN_TIMEOUT:
2895 tipc_sk(sk)->conn_timeout = value;
2897 case TIPC_MCAST_BROADCAST:
2898 tsk->mc_method.rcast = false;
2899 tsk->mc_method.mandatory = true;
2901 case TIPC_MCAST_REPLICAST:
2902 tsk->mc_method.rcast = true;
2903 tsk->mc_method.mandatory = true;
2905 case TIPC_GROUP_JOIN:
2906 res = tipc_sk_join(tsk, &mreq);
2908 case TIPC_GROUP_LEAVE:
2909 res = tipc_sk_leave(tsk);
2921 * tipc_getsockopt - get socket option
2922 * @sock: socket structure
2923 * @lvl: option level
2924 * @opt: option identifier
2925 * @ov: receptacle for option value
2926 * @ol: receptacle for length of option value
2928 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2929 * (to ease compatibility).
2931 * Returns 0 on success, errno otherwise
2933 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2934 char __user *ov, int __user *ol)
2936 struct sock *sk = sock->sk;
2937 struct tipc_sock *tsk = tipc_sk(sk);
2938 struct tipc_name_seq seq;
2943 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2944 return put_user(0, ol);
2945 if (lvl != SOL_TIPC)
2946 return -ENOPROTOOPT;
2947 res = get_user(len, ol);
2954 case TIPC_IMPORTANCE:
2955 value = tsk_importance(tsk);
2957 case TIPC_SRC_DROPPABLE:
2958 value = tsk_unreliable(tsk);
2960 case TIPC_DEST_DROPPABLE:
2961 value = tsk_unreturnable(tsk);
2963 case TIPC_CONN_TIMEOUT:
2964 value = tsk->conn_timeout;
2965 /* no need to set "res", since already 0 at this point */
2967 case TIPC_NODE_RECVQ_DEPTH:
2968 value = 0; /* was tipc_queue_size, now obsolete */
2970 case TIPC_SOCK_RECVQ_DEPTH:
2971 value = skb_queue_len(&sk->sk_receive_queue);
2973 case TIPC_GROUP_JOIN:
2976 tipc_group_self(tsk->group, &seq, &scope);
2986 return res; /* "get" failed */
2988 if (len < sizeof(value))
2991 if (copy_to_user(ov, &value, sizeof(value)))
2994 return put_user(sizeof(value), ol);
2997 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2999 struct net *net = sock_net(sock->sk);
3000 struct tipc_sioc_nodeid_req nr = {0};
3001 struct tipc_sioc_ln_req lnr;
3002 void __user *argp = (void __user *)arg;
3005 case SIOCGETLINKNAME:
3006 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3008 if (!tipc_node_get_linkname(net,
3009 lnr.bearer_id & 0xffff, lnr.peer,
3010 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3011 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3015 return -EADDRNOTAVAIL;
3017 if (copy_from_user(&nr, argp, sizeof(nr)))
3019 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3020 return -EADDRNOTAVAIL;
3021 if (copy_to_user(argp, &nr, sizeof(nr)))
3025 return -ENOIOCTLCMD;
3029 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3031 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3032 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3033 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3035 tsk1->peer.family = AF_TIPC;
3036 tsk1->peer.addrtype = TIPC_ADDR_ID;
3037 tsk1->peer.scope = TIPC_NODE_SCOPE;
3038 tsk1->peer.addr.id.ref = tsk2->portid;
3039 tsk1->peer.addr.id.node = onode;
3040 tsk2->peer.family = AF_TIPC;
3041 tsk2->peer.addrtype = TIPC_ADDR_ID;
3042 tsk2->peer.scope = TIPC_NODE_SCOPE;
3043 tsk2->peer.addr.id.ref = tsk1->portid;
3044 tsk2->peer.addr.id.node = onode;
3046 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3047 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3051 /* Protocol switches for the various types of TIPC sockets */
3053 static const struct proto_ops msg_ops = {
3054 .owner = THIS_MODULE,
3056 .release = tipc_release,
3058 .connect = tipc_connect,
3059 .socketpair = tipc_socketpair,
3060 .accept = sock_no_accept,
3061 .getname = tipc_getname,
3063 .ioctl = tipc_ioctl,
3064 .listen = sock_no_listen,
3065 .shutdown = tipc_shutdown,
3066 .setsockopt = tipc_setsockopt,
3067 .getsockopt = tipc_getsockopt,
3068 .sendmsg = tipc_sendmsg,
3069 .recvmsg = tipc_recvmsg,
3070 .mmap = sock_no_mmap,
3071 .sendpage = sock_no_sendpage
3074 static const struct proto_ops packet_ops = {
3075 .owner = THIS_MODULE,
3077 .release = tipc_release,
3079 .connect = tipc_connect,
3080 .socketpair = tipc_socketpair,
3081 .accept = tipc_accept,
3082 .getname = tipc_getname,
3084 .ioctl = tipc_ioctl,
3085 .listen = tipc_listen,
3086 .shutdown = tipc_shutdown,
3087 .setsockopt = tipc_setsockopt,
3088 .getsockopt = tipc_getsockopt,
3089 .sendmsg = tipc_send_packet,
3090 .recvmsg = tipc_recvmsg,
3091 .mmap = sock_no_mmap,
3092 .sendpage = sock_no_sendpage
3095 static const struct proto_ops stream_ops = {
3096 .owner = THIS_MODULE,
3098 .release = tipc_release,
3100 .connect = tipc_connect,
3101 .socketpair = tipc_socketpair,
3102 .accept = tipc_accept,
3103 .getname = tipc_getname,
3105 .ioctl = tipc_ioctl,
3106 .listen = tipc_listen,
3107 .shutdown = tipc_shutdown,
3108 .setsockopt = tipc_setsockopt,
3109 .getsockopt = tipc_getsockopt,
3110 .sendmsg = tipc_sendstream,
3111 .recvmsg = tipc_recvstream,
3112 .mmap = sock_no_mmap,
3113 .sendpage = sock_no_sendpage
3116 static const struct net_proto_family tipc_family_ops = {
3117 .owner = THIS_MODULE,
3119 .create = tipc_sk_create
3122 static struct proto tipc_proto = {
3124 .owner = THIS_MODULE,
3125 .obj_size = sizeof(struct tipc_sock),
3126 .sysctl_rmem = sysctl_tipc_rmem
3130 * tipc_socket_init - initialize TIPC socket interface
3132 * Returns 0 on success, errno otherwise
3134 int tipc_socket_init(void)
3138 res = proto_register(&tipc_proto, 1);
3140 pr_err("Failed to register TIPC protocol type\n");
3144 res = sock_register(&tipc_family_ops);
3146 pr_err("Failed to register TIPC socket type\n");
3147 proto_unregister(&tipc_proto);
3155 * tipc_socket_stop - stop TIPC socket interface
3157 void tipc_socket_stop(void)
3159 sock_unregister(tipc_family_ops.family);
3160 proto_unregister(&tipc_proto);
3163 /* Caller should hold socket lock for the passed tipc socket. */
3164 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3168 struct nlattr *nest;
3170 peer_node = tsk_peer_node(tsk);
3171 peer_port = tsk_peer_port(tsk);
3173 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3175 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3177 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3180 if (tsk->conn_type != 0) {
3181 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3183 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3185 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3188 nla_nest_end(skb, nest);
3193 nla_nest_cancel(skb, nest);
3198 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3201 struct net *net = sock_net(skb->sk);
3202 struct sock *sk = &tsk->sk;
3204 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3205 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3208 if (tipc_sk_connected(sk)) {
3209 if (__tipc_nl_add_sk_con(skb, tsk))
3211 } else if (!list_empty(&tsk->publications)) {
3212 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3218 /* Caller should hold socket lock for the passed tipc socket. */
3219 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3220 struct tipc_sock *tsk)
3222 struct nlattr *attrs;
3225 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3226 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3230 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3232 goto genlmsg_cancel;
3234 if (__tipc_nl_add_sk_info(skb, tsk))
3235 goto attr_msg_cancel;
3237 nla_nest_end(skb, attrs);
3238 genlmsg_end(skb, hdr);
3243 nla_nest_cancel(skb, attrs);
3245 genlmsg_cancel(skb, hdr);
3250 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3251 int (*skb_handler)(struct sk_buff *skb,
3252 struct netlink_callback *cb,
3253 struct tipc_sock *tsk))
3255 struct rhashtable_iter *iter = (void *)cb->args[4];
3256 struct tipc_sock *tsk;
3259 rhashtable_walk_start(iter);
3260 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3263 if (err == -EAGAIN) {
3270 sock_hold(&tsk->sk);
3271 rhashtable_walk_stop(iter);
3272 lock_sock(&tsk->sk);
3273 err = skb_handler(skb, cb, tsk);
3275 release_sock(&tsk->sk);
3279 release_sock(&tsk->sk);
3280 rhashtable_walk_start(iter);
3283 rhashtable_walk_stop(iter);
3287 EXPORT_SYMBOL(tipc_nl_sk_walk);
3289 int tipc_dump_start(struct netlink_callback *cb)
3291 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3293 EXPORT_SYMBOL(tipc_dump_start);
3295 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3297 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3298 struct rhashtable_iter *iter = (void *)cb->args[4];
3299 struct tipc_net *tn = tipc_net(net);
3302 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3306 cb->args[4] = (long)iter;
3309 rhashtable_walk_enter(&tn->sk_rht, iter);
3313 int tipc_dump_done(struct netlink_callback *cb)
3315 struct rhashtable_iter *hti = (void *)cb->args[4];
3317 rhashtable_walk_exit(hti);
3321 EXPORT_SYMBOL(tipc_dump_done);
3323 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3324 struct tipc_sock *tsk, u32 sk_filter_state,
3325 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3327 struct sock *sk = &tsk->sk;
3328 struct nlattr *attrs;
3329 struct nlattr *stat;
3331 /*filter response w.r.t sk_state*/
3332 if (!(sk_filter_state & (1 << sk->sk_state)))
3335 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3339 if (__tipc_nl_add_sk_info(skb, tsk))
3340 goto attr_msg_cancel;
3342 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3343 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3344 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3345 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3346 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3348 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3349 tipc_diag_gen_cookie(sk),
3351 goto attr_msg_cancel;
3353 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3355 goto attr_msg_cancel;
3357 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3358 skb_queue_len(&sk->sk_receive_queue)) ||
3359 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3360 skb_queue_len(&sk->sk_write_queue)) ||
3361 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3362 atomic_read(&sk->sk_drops)))
3363 goto stat_msg_cancel;
3365 if (tsk->cong_link_cnt &&
3366 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3367 goto stat_msg_cancel;
3369 if (tsk_conn_cong(tsk) &&
3370 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3371 goto stat_msg_cancel;
3373 nla_nest_end(skb, stat);
3376 if (tipc_group_fill_sock_diag(tsk->group, skb))
3377 goto stat_msg_cancel;
3379 nla_nest_end(skb, attrs);
3384 nla_nest_cancel(skb, stat);
3386 nla_nest_cancel(skb, attrs);
3390 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3392 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3394 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3397 /* Caller should hold socket lock for the passed tipc socket. */
3398 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3399 struct netlink_callback *cb,
3400 struct publication *publ)
3403 struct nlattr *attrs;
3405 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3406 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3410 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3412 goto genlmsg_cancel;
3414 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3415 goto attr_msg_cancel;
3416 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3417 goto attr_msg_cancel;
3418 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3419 goto attr_msg_cancel;
3420 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3421 goto attr_msg_cancel;
3423 nla_nest_end(skb, attrs);
3424 genlmsg_end(skb, hdr);
3429 nla_nest_cancel(skb, attrs);
3431 genlmsg_cancel(skb, hdr);
3436 /* Caller should hold socket lock for the passed tipc socket. */
3437 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3438 struct netlink_callback *cb,
3439 struct tipc_sock *tsk, u32 *last_publ)
3442 struct publication *p;
3445 list_for_each_entry(p, &tsk->publications, binding_sock) {
3446 if (p->key == *last_publ)
3449 if (p->key != *last_publ) {
3450 /* We never set seq or call nl_dump_check_consistent()
3451 * this means that setting prev_seq here will cause the
3452 * consistence check to fail in the netlink callback
3453 * handler. Resulting in the last NLMSG_DONE message
3454 * having the NLM_F_DUMP_INTR flag set.
3461 p = list_first_entry(&tsk->publications, struct publication,
3465 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3466 err = __tipc_nl_add_sk_publ(skb, cb, p);
3468 *last_publ = p->key;
3477 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3480 u32 tsk_portid = cb->args[0];
3481 u32 last_publ = cb->args[1];
3482 u32 done = cb->args[2];
3483 struct net *net = sock_net(skb->sk);
3484 struct tipc_sock *tsk;
3487 struct nlattr **attrs;
3488 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3490 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3494 if (!attrs[TIPC_NLA_SOCK])
3497 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3498 attrs[TIPC_NLA_SOCK],
3499 tipc_nl_sock_policy, NULL);
3503 if (!sock[TIPC_NLA_SOCK_REF])
3506 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3512 tsk = tipc_sk_lookup(net, tsk_portid);
3516 lock_sock(&tsk->sk);
3517 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3520 release_sock(&tsk->sk);
3523 cb->args[0] = tsk_portid;
3524 cb->args[1] = last_publ;