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"
51 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
52 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
53 #define TIPC_FWD_MSG 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
56 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
59 TIPC_LISTEN = TCP_LISTEN,
60 TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 TIPC_OPEN = TCP_CLOSE,
62 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 TIPC_CONNECTING = TCP_SYN_SENT,
66 struct sockaddr_pair {
67 struct sockaddr_tipc sock;
68 struct sockaddr_tipc member;
72 * struct tipc_sock - TIPC socket structure
73 * @sk: socket - interacts with 'port' and with user via the socket API
74 * @conn_type: TIPC type used when connection was established
75 * @conn_instance: TIPC instance used when connection was established
76 * @published: non-zero if port has one or more associated names
77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
78 * @maxnagle: maximum size of msg which can be subject to nagle
79 * @portid: unique port identity in TIPC socket hash table
80 * @phdr: preformatted message header used when sending messages
81 * #cong_links: list of congested links
82 * @publications: list of publications for port
83 * @blocking_link: address of the congested link we are currently sleeping on
84 * @pub_count: total # of publications port has made during its lifetime
85 * @conn_timeout: the time we can wait for an unresponded setup request
86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87 * @cong_link_cnt: number of congested links
88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
89 * @rcv_unacked: # messages read by user, but not yet acked back to peer
90 * @peer: 'connected' peer for dgram/rdm
91 * @node: hash table node
92 * @mc_method: cookie for use between socket and broadcast layer
93 * @rcu: rcu struct for tipc_sock
103 struct tipc_msg phdr;
104 struct list_head cong_links;
105 struct list_head publications;
107 atomic_t dupl_rcvcnt;
116 struct sockaddr_tipc peer;
117 struct rhash_head node;
118 struct tipc_mc_method mc_method;
120 struct tipc_group *group;
128 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
129 static void tipc_data_ready(struct sock *sk);
130 static void tipc_write_space(struct sock *sk);
131 static void tipc_sock_destruct(struct sock *sk);
132 static int tipc_release(struct socket *sock);
133 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
135 static void tipc_sk_timeout(struct timer_list *t);
136 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
137 struct tipc_name_seq const *seq);
138 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
139 struct tipc_name_seq const *seq);
140 static int tipc_sk_leave(struct tipc_sock *tsk);
141 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
142 static int tipc_sk_insert(struct tipc_sock *tsk);
143 static void tipc_sk_remove(struct tipc_sock *tsk);
144 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
145 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
146 static void tipc_sk_push_backlog(struct tipc_sock *tsk);
148 static const struct proto_ops packet_ops;
149 static const struct proto_ops stream_ops;
150 static const struct proto_ops msg_ops;
151 static struct proto tipc_proto;
152 static const struct rhashtable_params tsk_rht_params;
154 static u32 tsk_own_node(struct tipc_sock *tsk)
156 return msg_prevnode(&tsk->phdr);
159 static u32 tsk_peer_node(struct tipc_sock *tsk)
161 return msg_destnode(&tsk->phdr);
164 static u32 tsk_peer_port(struct tipc_sock *tsk)
166 return msg_destport(&tsk->phdr);
169 static bool tsk_unreliable(struct tipc_sock *tsk)
171 return msg_src_droppable(&tsk->phdr) != 0;
174 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
176 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
179 static bool tsk_unreturnable(struct tipc_sock *tsk)
181 return msg_dest_droppable(&tsk->phdr) != 0;
184 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
186 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
189 static int tsk_importance(struct tipc_sock *tsk)
191 return msg_importance(&tsk->phdr);
194 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
196 if (imp > TIPC_CRITICAL_IMPORTANCE)
198 msg_set_importance(&tsk->phdr, (u32)imp);
202 static struct tipc_sock *tipc_sk(const struct sock *sk)
204 return container_of(sk, struct tipc_sock, sk);
207 static bool tsk_conn_cong(struct tipc_sock *tsk)
209 return tsk->snt_unacked > tsk->snd_win;
212 static u16 tsk_blocks(int len)
214 return ((len / FLOWCTL_BLK_SZ) + 1);
217 /* tsk_blocks(): translate a buffer size in bytes to number of
218 * advertisable blocks, taking into account the ratio truesize(len)/len
219 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
221 static u16 tsk_adv_blocks(int len)
223 return len / FLOWCTL_BLK_SZ / 4;
226 /* tsk_inc(): increment counter for sent or received data
227 * - If block based flow control is not supported by peer we
228 * fall back to message based ditto, incrementing the counter
230 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
232 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
233 return ((msglen / FLOWCTL_BLK_SZ) + 1);
237 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
239 static void tsk_set_nagle(struct tipc_sock *tsk)
241 struct sock *sk = &tsk->sk;
244 if (sk->sk_type != SOCK_STREAM)
248 if (!(tsk->peer_caps & TIPC_NAGLE))
250 /* Limit node local buffer size to avoid receive queue overflow */
251 if (tsk->max_pkt == MAX_MSG_SIZE)
252 tsk->maxnagle = 1500;
254 tsk->maxnagle = tsk->max_pkt;
258 * tsk_advance_rx_queue - discard first buffer in socket receive queue
260 * Caller must hold socket lock
262 static void tsk_advance_rx_queue(struct sock *sk)
264 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
265 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
268 /* tipc_sk_respond() : send response message back to sender
270 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
274 u32 onode = tipc_own_addr(sock_net(sk));
276 if (!tipc_msg_reverse(onode, &skb, err))
279 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
280 dnode = msg_destnode(buf_msg(skb));
281 selector = msg_origport(buf_msg(skb));
282 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
286 * tsk_rej_rx_queue - reject all buffers in socket receive queue
288 * Caller must hold socket lock
290 static void tsk_rej_rx_queue(struct sock *sk)
294 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
295 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
298 static bool tipc_sk_connected(struct sock *sk)
300 return sk->sk_state == TIPC_ESTABLISHED;
303 /* tipc_sk_type_connectionless - check if the socket is datagram socket
306 * Returns true if connection less, false otherwise
308 static bool tipc_sk_type_connectionless(struct sock *sk)
310 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
313 /* tsk_peer_msg - verify if message was sent by connected port's peer
315 * Handles cases where the node's network address has changed from
316 * the default of <0.0.0> to its configured setting.
318 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
320 struct sock *sk = &tsk->sk;
321 u32 self = tipc_own_addr(sock_net(sk));
322 u32 peer_port = tsk_peer_port(tsk);
323 u32 orig_node, peer_node;
325 if (unlikely(!tipc_sk_connected(sk)))
328 if (unlikely(msg_origport(msg) != peer_port))
331 orig_node = msg_orignode(msg);
332 peer_node = tsk_peer_node(tsk);
334 if (likely(orig_node == peer_node))
337 if (!orig_node && peer_node == self)
340 if (!peer_node && orig_node == self)
346 /* tipc_set_sk_state - set the sk_state of the socket
349 * Caller must hold socket lock
351 * Returns 0 on success, errno otherwise
353 static int tipc_set_sk_state(struct sock *sk, int state)
355 int oldsk_state = sk->sk_state;
363 case TIPC_CONNECTING:
364 if (oldsk_state == TIPC_OPEN)
367 case TIPC_ESTABLISHED:
368 if (oldsk_state == TIPC_CONNECTING ||
369 oldsk_state == TIPC_OPEN)
372 case TIPC_DISCONNECTING:
373 if (oldsk_state == TIPC_CONNECTING ||
374 oldsk_state == TIPC_ESTABLISHED)
380 sk->sk_state = state;
385 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
387 struct sock *sk = sock->sk;
388 int err = sock_error(sk);
389 int typ = sock->type;
393 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
394 if (sk->sk_state == TIPC_DISCONNECTING)
396 else if (!tipc_sk_connected(sk))
401 if (signal_pending(current))
402 return sock_intr_errno(*timeout);
407 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
409 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
413 while ((rc_ = !(condition_))) { \
414 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
417 rc_ = tipc_sk_sock_err((sock_), timeo_); \
420 add_wait_queue(sk_sleep(sk_), &wait_); \
422 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
423 sched_annotate_sleep(); \
425 remove_wait_queue(sk_sleep(sk_), &wait_); \
431 * tipc_sk_create - create a TIPC socket
432 * @net: network namespace (must be default network)
433 * @sock: pre-allocated socket structure
434 * @protocol: protocol indicator (must be 0)
435 * @kern: caused by kernel or by userspace?
437 * This routine creates additional data structures used by the TIPC socket,
438 * initializes them, and links them together.
440 * Returns 0 on success, errno otherwise
442 static int tipc_sk_create(struct net *net, struct socket *sock,
443 int protocol, int kern)
445 const struct proto_ops *ops;
447 struct tipc_sock *tsk;
448 struct tipc_msg *msg;
450 /* Validate arguments */
451 if (unlikely(protocol != 0))
452 return -EPROTONOSUPPORT;
454 switch (sock->type) {
469 /* Allocate socket's protocol area */
470 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
475 tsk->max_pkt = MAX_PKT_DEFAULT;
477 INIT_LIST_HEAD(&tsk->publications);
478 INIT_LIST_HEAD(&tsk->cong_links);
481 /* Finish initializing socket data structures */
483 sock_init_data(sock, sk);
484 tipc_set_sk_state(sk, TIPC_OPEN);
485 if (tipc_sk_insert(tsk)) {
486 pr_warn("Socket create failed; port number exhausted\n");
490 /* Ensure tsk is visible before we read own_addr. */
493 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
494 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
496 msg_set_origport(msg, tsk->portid);
497 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
499 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
500 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
501 sk->sk_data_ready = tipc_data_ready;
502 sk->sk_write_space = tipc_write_space;
503 sk->sk_destruct = tipc_sock_destruct;
504 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
505 tsk->group_is_open = true;
506 atomic_set(&tsk->dupl_rcvcnt, 0);
508 /* Start out with safe limits until we receive an advertised window */
509 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
510 tsk->rcv_win = tsk->snd_win;
512 if (tipc_sk_type_connectionless(sk)) {
513 tsk_set_unreturnable(tsk, true);
514 if (sock->type == SOCK_DGRAM)
515 tsk_set_unreliable(tsk, true);
517 __skb_queue_head_init(&tsk->mc_method.deferredq);
518 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
522 static void tipc_sk_callback(struct rcu_head *head)
524 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
529 /* Caller should hold socket lock for the socket. */
530 static void __tipc_shutdown(struct socket *sock, int error)
532 struct sock *sk = sock->sk;
533 struct tipc_sock *tsk = tipc_sk(sk);
534 struct net *net = sock_net(sk);
535 long timeout = CONN_TIMEOUT_DEFAULT;
536 u32 dnode = tsk_peer_node(tsk);
539 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
540 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
541 !tsk_conn_cong(tsk)));
543 /* Push out unsent messages or remove if pending SYN */
544 skb = skb_peek(&sk->sk_write_queue);
545 if (skb && !msg_is_syn(buf_msg(skb)))
546 tipc_sk_push_backlog(tsk);
548 __skb_queue_purge(&sk->sk_write_queue);
550 /* Reject all unreceived messages, except on an active connection
551 * (which disconnects locally & sends a 'FIN+' to peer).
553 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
554 if (TIPC_SKB_CB(skb)->bytes_read) {
558 if (!tipc_sk_type_connectionless(sk) &&
559 sk->sk_state != TIPC_DISCONNECTING) {
560 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
561 tipc_node_remove_conn(net, dnode, tsk->portid);
563 tipc_sk_respond(sk, skb, error);
566 if (tipc_sk_type_connectionless(sk))
569 if (sk->sk_state != TIPC_DISCONNECTING) {
570 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
571 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
572 tsk_own_node(tsk), tsk_peer_port(tsk),
575 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
576 tipc_node_remove_conn(net, dnode, tsk->portid);
577 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
582 * tipc_release - destroy a TIPC socket
583 * @sock: socket to destroy
585 * This routine cleans up any messages that are still queued on the socket.
586 * For DGRAM and RDM socket types, all queued messages are rejected.
587 * For SEQPACKET and STREAM socket types, the first message is rejected
588 * and any others are discarded. (If the first message on a STREAM socket
589 * is partially-read, it is discarded and the next one is rejected instead.)
591 * NOTE: Rejected messages are not necessarily returned to the sender! They
592 * are returned or discarded according to the "destination droppable" setting
593 * specified for the message by the sender.
595 * Returns 0 on success, errno otherwise
597 static int tipc_release(struct socket *sock)
599 struct sock *sk = sock->sk;
600 struct tipc_sock *tsk;
603 * Exit if socket isn't fully initialized (occurs when a failed accept()
604 * releases a pre-allocated child socket that was never used)
612 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
613 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
614 sk->sk_shutdown = SHUTDOWN_MASK;
616 tipc_sk_withdraw(tsk, 0, NULL);
617 __skb_queue_purge(&tsk->mc_method.deferredq);
618 sk_stop_timer(sk, &sk->sk_timer);
622 /* Reject any messages that accumulated in backlog queue */
624 tipc_dest_list_purge(&tsk->cong_links);
625 tsk->cong_link_cnt = 0;
626 call_rcu(&tsk->rcu, tipc_sk_callback);
633 * tipc_bind - associate or disassocate TIPC name(s) with a socket
634 * @sock: socket structure
635 * @uaddr: socket address describing name(s) and desired operation
636 * @uaddr_len: size of socket address data structure
638 * Name and name sequence binding is indicated using a positive scope value;
639 * a negative scope value unbinds the specified name. Specifying no name
640 * (i.e. a socket address length of 0) unbinds all names from the socket.
642 * Returns 0 on success, errno otherwise
644 * NOTE: This routine doesn't need to take the socket lock since it doesn't
645 * access any non-constant socket information.
647 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
650 struct sock *sk = sock->sk;
651 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
652 struct tipc_sock *tsk = tipc_sk(sk);
656 if (unlikely(!uaddr_len)) {
657 res = tipc_sk_withdraw(tsk, 0, NULL);
664 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
668 if (addr->family != AF_TIPC) {
673 if (addr->addrtype == TIPC_ADDR_NAME)
674 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
675 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
680 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
681 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
682 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
687 res = (addr->scope >= 0) ?
688 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
689 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
696 * tipc_getname - get port ID of socket or peer socket
697 * @sock: socket structure
698 * @uaddr: area for returned socket address
699 * @uaddr_len: area for returned length of socket address
700 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
702 * Returns 0 on success, errno otherwise
704 * NOTE: This routine doesn't need to take the socket lock since it only
705 * accesses socket information that is unchanging (or which changes in
706 * a completely predictable manner).
708 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
711 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
712 struct sock *sk = sock->sk;
713 struct tipc_sock *tsk = tipc_sk(sk);
715 memset(addr, 0, sizeof(*addr));
717 if ((!tipc_sk_connected(sk)) &&
718 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
720 addr->addr.id.ref = tsk_peer_port(tsk);
721 addr->addr.id.node = tsk_peer_node(tsk);
723 addr->addr.id.ref = tsk->portid;
724 addr->addr.id.node = tipc_own_addr(sock_net(sk));
727 addr->addrtype = TIPC_ADDR_ID;
728 addr->family = AF_TIPC;
730 addr->addr.name.domain = 0;
732 return sizeof(*addr);
736 * tipc_poll - read and possibly block on pollmask
737 * @file: file structure associated with the socket
738 * @sock: socket for which to calculate the poll bits
741 * Returns pollmask value
744 * It appears that the usual socket locking mechanisms are not useful here
745 * since the pollmask info is potentially out-of-date the moment this routine
746 * exits. TCP and other protocols seem to rely on higher level poll routines
747 * to handle any preventable race conditions, so TIPC will do the same ...
749 * IMPORTANT: The fact that a read or write operation is indicated does NOT
750 * imply that the operation will succeed, merely that it should be performed
751 * and will not block.
753 static __poll_t tipc_poll(struct file *file, struct socket *sock,
756 struct sock *sk = sock->sk;
757 struct tipc_sock *tsk = tipc_sk(sk);
758 __poll_t revents = 0;
760 sock_poll_wait(file, sock, wait);
761 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
763 if (sk->sk_shutdown & RCV_SHUTDOWN)
764 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
765 if (sk->sk_shutdown == SHUTDOWN_MASK)
768 switch (sk->sk_state) {
769 case TIPC_ESTABLISHED:
770 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
774 case TIPC_CONNECTING:
775 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
776 revents |= EPOLLIN | EPOLLRDNORM;
779 if (tsk->group_is_open && !tsk->cong_link_cnt)
781 if (!tipc_sk_type_connectionless(sk))
783 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
785 revents |= EPOLLIN | EPOLLRDNORM;
787 case TIPC_DISCONNECTING:
788 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
795 * tipc_sendmcast - send multicast message
796 * @sock: socket structure
797 * @seq: destination address
798 * @msg: message to send
799 * @dlen: length of data to send
800 * @timeout: timeout to wait for wakeup
802 * Called from function tipc_sendmsg(), which has done all sanity checks
803 * Returns the number of bytes sent on success, or errno
805 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
806 struct msghdr *msg, size_t dlen, long timeout)
808 struct sock *sk = sock->sk;
809 struct tipc_sock *tsk = tipc_sk(sk);
810 struct tipc_msg *hdr = &tsk->phdr;
811 struct net *net = sock_net(sk);
812 int mtu = tipc_bcast_get_mtu(net);
813 struct tipc_mc_method *method = &tsk->mc_method;
814 struct sk_buff_head pkts;
815 struct tipc_nlist dsts;
821 /* Block or return if any destination link is congested */
822 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
826 /* Lookup destination nodes */
827 tipc_nlist_init(&dsts, tipc_own_addr(net));
828 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
830 if (!dsts.local && !dsts.remote)
831 return -EHOSTUNREACH;
833 /* Build message header */
834 msg_set_type(hdr, TIPC_MCAST_MSG);
835 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
836 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
837 msg_set_destport(hdr, 0);
838 msg_set_destnode(hdr, 0);
839 msg_set_nametype(hdr, seq->type);
840 msg_set_namelower(hdr, seq->lower);
841 msg_set_nameupper(hdr, seq->upper);
843 /* Build message as chain of buffers */
844 __skb_queue_head_init(&pkts);
845 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
847 /* Send message if build was successful */
848 if (unlikely(rc == dlen)) {
849 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
850 TIPC_DUMP_SK_SNDQ, " ");
851 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
852 &tsk->cong_link_cnt);
855 tipc_nlist_purge(&dsts);
857 return rc ? rc : dlen;
861 * tipc_send_group_msg - send a message to a member in the group
862 * @net: network namespace
863 * @m: message to send
865 * @dnode: destination node
866 * @dport: destination port
867 * @dlen: total length of message data
869 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
870 struct msghdr *m, struct tipc_member *mb,
871 u32 dnode, u32 dport, int dlen)
873 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
874 struct tipc_mc_method *method = &tsk->mc_method;
875 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
876 struct tipc_msg *hdr = &tsk->phdr;
877 struct sk_buff_head pkts;
880 /* Complete message header */
881 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
882 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
883 msg_set_destport(hdr, dport);
884 msg_set_destnode(hdr, dnode);
885 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
887 /* Build message as chain of buffers */
888 __skb_queue_head_init(&pkts);
889 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
890 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
891 if (unlikely(rc != dlen))
895 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
896 if (unlikely(rc == -ELINKCONG)) {
897 tipc_dest_push(&tsk->cong_links, dnode, 0);
898 tsk->cong_link_cnt++;
901 /* Update send window */
902 tipc_group_update_member(mb, blks);
904 /* A broadcast sent within next EXPIRE period must follow same path */
905 method->rcast = true;
906 method->mandatory = true;
911 * tipc_send_group_unicast - send message to a member in the group
912 * @sock: socket structure
913 * @m: message to send
914 * @dlen: total length of message data
915 * @timeout: timeout to wait for wakeup
917 * Called from function tipc_sendmsg(), which has done all sanity checks
918 * Returns the number of bytes sent on success, or errno
920 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
921 int dlen, long timeout)
923 struct sock *sk = sock->sk;
924 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
925 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
926 struct tipc_sock *tsk = tipc_sk(sk);
927 struct net *net = sock_net(sk);
928 struct tipc_member *mb = NULL;
932 node = dest->addr.id.node;
933 port = dest->addr.id.ref;
935 return -EHOSTUNREACH;
937 /* Block or return if destination link or member is congested */
938 rc = tipc_wait_for_cond(sock, &timeout,
939 !tipc_dest_find(&tsk->cong_links, node, 0) &&
941 !tipc_group_cong(tsk->group, node, port, blks,
947 return -EHOSTUNREACH;
949 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
951 return rc ? rc : dlen;
955 * tipc_send_group_anycast - send message to any member with given identity
956 * @sock: socket structure
957 * @m: message to send
958 * @dlen: total length of message data
959 * @timeout: timeout to wait for wakeup
961 * Called from function tipc_sendmsg(), which has done all sanity checks
962 * Returns the number of bytes sent on success, or errno
964 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
965 int dlen, long timeout)
967 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
968 struct sock *sk = sock->sk;
969 struct tipc_sock *tsk = tipc_sk(sk);
970 struct list_head *cong_links = &tsk->cong_links;
971 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
972 struct tipc_msg *hdr = &tsk->phdr;
973 struct tipc_member *first = NULL;
974 struct tipc_member *mbr = NULL;
975 struct net *net = sock_net(sk);
976 u32 node, port, exclude;
977 struct list_head dsts;
978 u32 type, inst, scope;
983 INIT_LIST_HEAD(&dsts);
985 type = msg_nametype(hdr);
986 inst = dest->addr.name.name.instance;
987 scope = msg_lookup_scope(hdr);
989 while (++lookups < 4) {
990 exclude = tipc_group_exclude(tsk->group);
994 /* Look for a non-congested destination member, if any */
996 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
997 &dstcnt, exclude, false))
998 return -EHOSTUNREACH;
999 tipc_dest_pop(&dsts, &node, &port);
1000 cong = tipc_group_cong(tsk->group, node, port, blks,
1010 /* Start over if destination was not in member list */
1014 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1017 /* Block or return if destination link or member is congested */
1018 rc = tipc_wait_for_cond(sock, &timeout,
1019 !tipc_dest_find(cong_links, node, 0) &&
1021 !tipc_group_cong(tsk->group, node, port,
1026 /* Send, unless destination disappeared while waiting */
1031 if (unlikely(lookups >= 4))
1032 return -EHOSTUNREACH;
1034 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1036 return rc ? rc : dlen;
1040 * tipc_send_group_bcast - send message to all members in communication group
1041 * @sk: socket structure
1042 * @m: message to send
1043 * @dlen: total length of message data
1044 * @timeout: timeout to wait for wakeup
1046 * Called from function tipc_sendmsg(), which has done all sanity checks
1047 * Returns the number of bytes sent on success, or errno
1049 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1050 int dlen, long timeout)
1052 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1053 struct sock *sk = sock->sk;
1054 struct net *net = sock_net(sk);
1055 struct tipc_sock *tsk = tipc_sk(sk);
1056 struct tipc_nlist *dsts;
1057 struct tipc_mc_method *method = &tsk->mc_method;
1058 bool ack = method->mandatory && method->rcast;
1059 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1060 struct tipc_msg *hdr = &tsk->phdr;
1061 int mtu = tipc_bcast_get_mtu(net);
1062 struct sk_buff_head pkts;
1063 int rc = -EHOSTUNREACH;
1065 /* Block or return if any destination link or member is congested */
1066 rc = tipc_wait_for_cond(sock, &timeout,
1067 !tsk->cong_link_cnt && tsk->group &&
1068 !tipc_group_bc_cong(tsk->group, blks));
1072 dsts = tipc_group_dests(tsk->group);
1073 if (!dsts->local && !dsts->remote)
1074 return -EHOSTUNREACH;
1076 /* Complete message header */
1078 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1079 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1081 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1082 msg_set_nameinst(hdr, 0);
1084 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1085 msg_set_destport(hdr, 0);
1086 msg_set_destnode(hdr, 0);
1087 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1089 /* Avoid getting stuck with repeated forced replicasts */
1090 msg_set_grp_bc_ack_req(hdr, ack);
1092 /* Build message as chain of buffers */
1093 __skb_queue_head_init(&pkts);
1094 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1095 if (unlikely(rc != dlen))
1099 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1103 /* Update broadcast sequence number and send windows */
1104 tipc_group_update_bc_members(tsk->group, blks, ack);
1106 /* Broadcast link is now free to choose method for next broadcast */
1107 method->mandatory = false;
1108 method->expires = jiffies;
1114 * tipc_send_group_mcast - send message to all members with given identity
1115 * @sock: socket structure
1116 * @m: message to send
1117 * @dlen: total length of message data
1118 * @timeout: timeout to wait for wakeup
1120 * Called from function tipc_sendmsg(), which has done all sanity checks
1121 * Returns the number of bytes sent on success, or errno
1123 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1124 int dlen, long timeout)
1126 struct sock *sk = sock->sk;
1127 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1128 struct tipc_sock *tsk = tipc_sk(sk);
1129 struct tipc_group *grp = tsk->group;
1130 struct tipc_msg *hdr = &tsk->phdr;
1131 struct net *net = sock_net(sk);
1132 u32 type, inst, scope, exclude;
1133 struct list_head dsts;
1136 INIT_LIST_HEAD(&dsts);
1138 type = msg_nametype(hdr);
1139 inst = dest->addr.name.name.instance;
1140 scope = msg_lookup_scope(hdr);
1141 exclude = tipc_group_exclude(grp);
1143 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1144 &dstcnt, exclude, true))
1145 return -EHOSTUNREACH;
1148 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1149 return tipc_send_group_unicast(sock, m, dlen, timeout);
1152 tipc_dest_list_purge(&dsts);
1153 return tipc_send_group_bcast(sock, m, dlen, timeout);
1157 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1158 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1159 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1161 * Multi-threaded: parallel calls with reference to same queues may occur
1163 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1164 struct sk_buff_head *inputq)
1166 u32 self = tipc_own_addr(net);
1167 u32 type, lower, upper, scope;
1168 struct sk_buff *skb, *_skb;
1170 struct sk_buff_head tmpq;
1171 struct list_head dports;
1172 struct tipc_msg *hdr;
1173 int user, mtyp, hlen;
1176 __skb_queue_head_init(&tmpq);
1177 INIT_LIST_HEAD(&dports);
1179 skb = tipc_skb_peek(arrvq, &inputq->lock);
1180 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1182 user = msg_user(hdr);
1183 mtyp = msg_type(hdr);
1184 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1185 onode = msg_orignode(hdr);
1186 type = msg_nametype(hdr);
1188 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1189 spin_lock_bh(&inputq->lock);
1190 if (skb_peek(arrvq) == skb) {
1191 __skb_dequeue(arrvq);
1192 __skb_queue_tail(inputq, skb);
1195 spin_unlock_bh(&inputq->lock);
1199 /* Group messages require exact scope match */
1200 if (msg_in_group(hdr)) {
1203 scope = msg_lookup_scope(hdr);
1206 /* TIPC_NODE_SCOPE means "any scope" in this context */
1208 scope = TIPC_NODE_SCOPE;
1210 scope = TIPC_CLUSTER_SCOPE;
1212 lower = msg_namelower(hdr);
1213 upper = msg_nameupper(hdr);
1216 /* Create destination port list: */
1217 tipc_nametbl_mc_lookup(net, type, lower, upper,
1218 scope, exact, &dports);
1220 /* Clone message per destination */
1221 while (tipc_dest_pop(&dports, NULL, &portid)) {
1222 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1224 msg_set_destport(buf_msg(_skb), portid);
1225 __skb_queue_tail(&tmpq, _skb);
1228 pr_warn("Failed to clone mcast rcv buffer\n");
1230 /* Append to inputq if not already done by other thread */
1231 spin_lock_bh(&inputq->lock);
1232 if (skb_peek(arrvq) == skb) {
1233 skb_queue_splice_tail_init(&tmpq, inputq);
1234 kfree_skb(__skb_dequeue(arrvq));
1236 spin_unlock_bh(&inputq->lock);
1237 __skb_queue_purge(&tmpq);
1240 tipc_sk_rcv(net, inputq);
1243 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1244 * when socket is in Nagle mode
1246 static void tipc_sk_push_backlog(struct tipc_sock *tsk)
1248 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1249 struct net *net = sock_net(&tsk->sk);
1250 u32 dnode = tsk_peer_node(tsk);
1253 if (skb_queue_empty(txq) || tsk->cong_link_cnt)
1256 tsk->snt_unacked += tsk->snd_backlog;
1257 tsk->snd_backlog = 0;
1258 tsk->expect_ack = true;
1259 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1260 if (rc == -ELINKCONG)
1261 tsk->cong_link_cnt = 1;
1265 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1266 * @tsk: receiving socket
1267 * @skb: pointer to message buffer.
1269 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1270 struct sk_buff_head *inputq,
1271 struct sk_buff_head *xmitq)
1273 struct tipc_msg *hdr = buf_msg(skb);
1274 u32 onode = tsk_own_node(tsk);
1275 struct sock *sk = &tsk->sk;
1276 int mtyp = msg_type(hdr);
1279 /* Ignore if connection cannot be validated: */
1280 if (!tsk_peer_msg(tsk, hdr)) {
1281 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1285 if (unlikely(msg_errcode(hdr))) {
1286 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1287 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1288 tsk_peer_port(tsk));
1289 sk->sk_state_change(sk);
1291 /* State change is ignored if socket already awake,
1292 * - convert msg to abort msg and add to inqueue
1294 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1295 msg_set_type(hdr, TIPC_CONN_MSG);
1296 msg_set_size(hdr, BASIC_H_SIZE);
1297 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1298 __skb_queue_tail(inputq, skb);
1302 tsk->probe_unacked = false;
1304 if (mtyp == CONN_PROBE) {
1305 msg_set_type(hdr, CONN_PROBE_REPLY);
1306 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1307 __skb_queue_tail(xmitq, skb);
1309 } else if (mtyp == CONN_ACK) {
1310 was_cong = tsk_conn_cong(tsk);
1311 tsk->expect_ack = false;
1312 tipc_sk_push_backlog(tsk);
1313 tsk->snt_unacked -= msg_conn_ack(hdr);
1314 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1315 tsk->snd_win = msg_adv_win(hdr);
1316 if (was_cong && !tsk_conn_cong(tsk))
1317 sk->sk_write_space(sk);
1318 } else if (mtyp != CONN_PROBE_REPLY) {
1319 pr_warn("Received unknown CONN_PROTO msg\n");
1326 * tipc_sendmsg - send message in connectionless manner
1327 * @sock: socket structure
1328 * @m: message to send
1329 * @dsz: amount of user data to be sent
1331 * Message must have an destination specified explicitly.
1332 * Used for SOCK_RDM and SOCK_DGRAM messages,
1333 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1334 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1336 * Returns the number of bytes sent on success, or errno otherwise
1338 static int tipc_sendmsg(struct socket *sock,
1339 struct msghdr *m, size_t dsz)
1341 struct sock *sk = sock->sk;
1345 ret = __tipc_sendmsg(sock, m, dsz);
1351 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1353 struct sock *sk = sock->sk;
1354 struct net *net = sock_net(sk);
1355 struct tipc_sock *tsk = tipc_sk(sk);
1356 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1357 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1358 struct list_head *clinks = &tsk->cong_links;
1359 bool syn = !tipc_sk_type_connectionless(sk);
1360 struct tipc_group *grp = tsk->group;
1361 struct tipc_msg *hdr = &tsk->phdr;
1362 struct tipc_name_seq *seq;
1363 struct sk_buff_head pkts;
1364 u32 dport, dnode = 0;
1368 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1372 if (unlikely(m->msg_namelen < sizeof(*dest)))
1374 if (unlikely(dest->family != AF_TIPC))
1380 return tipc_send_group_bcast(sock, m, dlen, timeout);
1381 if (dest->addrtype == TIPC_ADDR_NAME)
1382 return tipc_send_group_anycast(sock, m, dlen, timeout);
1383 if (dest->addrtype == TIPC_ADDR_ID)
1384 return tipc_send_group_unicast(sock, m, dlen, timeout);
1385 if (dest->addrtype == TIPC_ADDR_MCAST)
1386 return tipc_send_group_mcast(sock, m, dlen, timeout);
1390 if (unlikely(!dest)) {
1392 if (!syn && dest->family != AF_TIPC)
1393 return -EDESTADDRREQ;
1396 if (unlikely(syn)) {
1397 if (sk->sk_state == TIPC_LISTEN)
1399 if (sk->sk_state != TIPC_OPEN)
1403 if (dest->addrtype == TIPC_ADDR_NAME) {
1404 tsk->conn_type = dest->addr.name.name.type;
1405 tsk->conn_instance = dest->addr.name.name.instance;
1407 msg_set_syn(hdr, 1);
1410 seq = &dest->addr.nameseq;
1411 if (dest->addrtype == TIPC_ADDR_MCAST)
1412 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1414 if (dest->addrtype == TIPC_ADDR_NAME) {
1415 type = dest->addr.name.name.type;
1416 inst = dest->addr.name.name.instance;
1417 dnode = dest->addr.name.domain;
1418 msg_set_type(hdr, TIPC_NAMED_MSG);
1419 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1420 msg_set_nametype(hdr, type);
1421 msg_set_nameinst(hdr, inst);
1422 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1423 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1424 msg_set_destnode(hdr, dnode);
1425 msg_set_destport(hdr, dport);
1426 if (unlikely(!dport && !dnode))
1427 return -EHOSTUNREACH;
1428 } else if (dest->addrtype == TIPC_ADDR_ID) {
1429 dnode = dest->addr.id.node;
1430 msg_set_type(hdr, TIPC_DIRECT_MSG);
1431 msg_set_lookup_scope(hdr, 0);
1432 msg_set_destnode(hdr, dnode);
1433 msg_set_destport(hdr, dest->addr.id.ref);
1434 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1439 /* Block or return if destination link is congested */
1440 rc = tipc_wait_for_cond(sock, &timeout,
1441 !tipc_dest_find(clinks, dnode, 0));
1445 __skb_queue_head_init(&pkts);
1446 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
1447 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1448 if (unlikely(rc != dlen))
1450 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1453 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1454 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1455 if (unlikely(rc == -ELINKCONG)) {
1456 tipc_dest_push(clinks, dnode, 0);
1457 tsk->cong_link_cnt++;
1461 if (unlikely(syn && !rc))
1462 tipc_set_sk_state(sk, TIPC_CONNECTING);
1464 return rc ? rc : dlen;
1468 * tipc_sendstream - send stream-oriented data
1469 * @sock: socket structure
1471 * @dsz: total length of data to be transmitted
1473 * Used for SOCK_STREAM data.
1475 * Returns the number of bytes sent on success (or partial success),
1476 * or errno if no data sent
1478 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1480 struct sock *sk = sock->sk;
1484 ret = __tipc_sendstream(sock, m, dsz);
1490 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1492 struct sock *sk = sock->sk;
1493 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1494 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1495 struct sk_buff_head *txq = &sk->sk_write_queue;
1496 struct tipc_sock *tsk = tipc_sk(sk);
1497 struct tipc_msg *hdr = &tsk->phdr;
1498 struct net *net = sock_net(sk);
1499 u32 dnode = tsk_peer_node(tsk);
1500 int maxnagle = tsk->maxnagle;
1501 int maxpkt = tsk->max_pkt;
1505 if (unlikely(dlen > INT_MAX))
1508 /* Handle implicit connection setup */
1509 if (unlikely(dest)) {
1510 rc = __tipc_sendmsg(sock, m, dlen);
1511 if (dlen && dlen == rc) {
1512 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1513 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1519 rc = tipc_wait_for_cond(sock, &timeout,
1520 (!tsk->cong_link_cnt &&
1521 !tsk_conn_cong(tsk) &&
1522 tipc_sk_connected(sk)));
1525 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1526 blocks = tsk->snd_backlog;
1527 if (tsk->oneway++ >= 4 && send <= maxnagle) {
1528 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1529 if (unlikely(rc < 0))
1532 if (blocks <= 64 && tsk->expect_ack) {
1533 tsk->snd_backlog = blocks;
1537 tsk->expect_ack = true;
1539 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1540 if (unlikely(rc != send))
1542 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1544 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1545 TIPC_DUMP_SK_SNDQ, " ");
1546 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1547 if (unlikely(rc == -ELINKCONG)) {
1548 tsk->cong_link_cnt = 1;
1552 tsk->snt_unacked += blocks;
1553 tsk->snd_backlog = 0;
1556 } while (sent < dlen && !rc);
1558 return sent ? sent : rc;
1562 * tipc_send_packet - send a connection-oriented message
1563 * @sock: socket structure
1564 * @m: message to send
1565 * @dsz: length of data to be transmitted
1567 * Used for SOCK_SEQPACKET messages.
1569 * Returns the number of bytes sent on success, or errno otherwise
1571 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1573 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1576 return tipc_sendstream(sock, m, dsz);
1579 /* tipc_sk_finish_conn - complete the setup of a connection
1581 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1584 struct sock *sk = &tsk->sk;
1585 struct net *net = sock_net(sk);
1586 struct tipc_msg *msg = &tsk->phdr;
1588 msg_set_syn(msg, 0);
1589 msg_set_destnode(msg, peer_node);
1590 msg_set_destport(msg, peer_port);
1591 msg_set_type(msg, TIPC_CONN_MSG);
1592 msg_set_lookup_scope(msg, 0);
1593 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1595 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1596 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1597 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1598 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1599 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1601 __skb_queue_purge(&sk->sk_write_queue);
1602 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1605 /* Fall back to message based flow control */
1606 tsk->rcv_win = FLOWCTL_MSG_WIN;
1607 tsk->snd_win = FLOWCTL_MSG_WIN;
1611 * tipc_sk_set_orig_addr - capture sender's address for received message
1612 * @m: descriptor for message info
1613 * @hdr: received message header
1615 * Note: Address is not captured if not requested by receiver.
1617 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1619 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1620 struct tipc_msg *hdr = buf_msg(skb);
1625 srcaddr->sock.family = AF_TIPC;
1626 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1627 srcaddr->sock.scope = 0;
1628 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1629 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1630 srcaddr->sock.addr.name.domain = 0;
1631 m->msg_namelen = sizeof(struct sockaddr_tipc);
1633 if (!msg_in_group(hdr))
1636 /* Group message users may also want to know sending member's id */
1637 srcaddr->member.family = AF_TIPC;
1638 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1639 srcaddr->member.scope = 0;
1640 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1641 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1642 srcaddr->member.addr.name.domain = 0;
1643 m->msg_namelen = sizeof(*srcaddr);
1647 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1648 * @m: descriptor for message info
1649 * @skb: received message buffer
1650 * @tsk: TIPC port associated with message
1652 * Note: Ancillary data is not captured if not requested by receiver.
1654 * Returns 0 if successful, otherwise errno
1656 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1657 struct tipc_sock *tsk)
1659 struct tipc_msg *msg;
1666 if (likely(m->msg_controllen == 0))
1670 /* Optionally capture errored message object(s) */
1671 err = msg ? msg_errcode(msg) : 0;
1672 if (unlikely(err)) {
1674 anc_data[1] = msg_data_sz(msg);
1675 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1679 if (skb_linearize(skb))
1682 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1689 /* Optionally capture message destination object */
1690 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1691 switch (dest_type) {
1692 case TIPC_NAMED_MSG:
1694 anc_data[0] = msg_nametype(msg);
1695 anc_data[1] = msg_namelower(msg);
1696 anc_data[2] = msg_namelower(msg);
1698 case TIPC_MCAST_MSG:
1700 anc_data[0] = msg_nametype(msg);
1701 anc_data[1] = msg_namelower(msg);
1702 anc_data[2] = msg_nameupper(msg);
1705 has_name = (tsk->conn_type != 0);
1706 anc_data[0] = tsk->conn_type;
1707 anc_data[1] = tsk->conn_instance;
1708 anc_data[2] = tsk->conn_instance;
1714 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1722 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1724 struct sock *sk = &tsk->sk;
1725 struct net *net = sock_net(sk);
1726 struct sk_buff *skb = NULL;
1727 struct tipc_msg *msg;
1728 u32 peer_port = tsk_peer_port(tsk);
1729 u32 dnode = tsk_peer_node(tsk);
1731 if (!tipc_sk_connected(sk))
1733 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1734 dnode, tsk_own_node(tsk), peer_port,
1735 tsk->portid, TIPC_OK);
1739 msg_set_conn_ack(msg, tsk->rcv_unacked);
1740 tsk->rcv_unacked = 0;
1742 /* Adjust to and advertize the correct window limit */
1743 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1744 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1745 msg_set_adv_win(msg, tsk->rcv_win);
1747 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1750 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1752 struct sock *sk = sock->sk;
1753 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1754 long timeo = *timeop;
1755 int err = sock_error(sk);
1761 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1762 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1766 add_wait_queue(sk_sleep(sk), &wait);
1768 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1769 sched_annotate_sleep();
1771 remove_wait_queue(sk_sleep(sk), &wait);
1774 if (!skb_queue_empty(&sk->sk_receive_queue))
1779 err = sock_intr_errno(timeo);
1780 if (signal_pending(current))
1783 err = sock_error(sk);
1792 * tipc_recvmsg - receive packet-oriented message
1793 * @m: descriptor for message info
1794 * @buflen: length of user buffer area
1795 * @flags: receive flags
1797 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1798 * If the complete message doesn't fit in user area, truncate it.
1800 * Returns size of returned message data, errno otherwise
1802 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1803 size_t buflen, int flags)
1805 struct sock *sk = sock->sk;
1806 bool connected = !tipc_sk_type_connectionless(sk);
1807 struct tipc_sock *tsk = tipc_sk(sk);
1808 int rc, err, hlen, dlen, copy;
1809 struct sk_buff_head xmitq;
1810 struct tipc_msg *hdr;
1811 struct sk_buff *skb;
1815 /* Catch invalid receive requests */
1816 if (unlikely(!buflen))
1820 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1824 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1826 /* Step rcv queue to first msg with data or error; wait if necessary */
1828 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1831 skb = skb_peek(&sk->sk_receive_queue);
1833 dlen = msg_data_sz(hdr);
1834 hlen = msg_hdr_sz(hdr);
1835 err = msg_errcode(hdr);
1836 grp_evt = msg_is_grp_evt(hdr);
1837 if (likely(dlen || err))
1839 tsk_advance_rx_queue(sk);
1842 /* Collect msg meta data, including error code and rejected data */
1843 tipc_sk_set_orig_addr(m, skb);
1844 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1849 /* Capture data if non-error msg, otherwise just set return value */
1851 copy = min_t(int, dlen, buflen);
1852 if (unlikely(copy != dlen))
1853 m->msg_flags |= MSG_TRUNC;
1854 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1858 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1864 /* Mark message as group event if applicable */
1865 if (unlikely(grp_evt)) {
1866 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1867 m->msg_flags |= MSG_EOR;
1868 m->msg_flags |= MSG_OOB;
1872 /* Caption of data or error code/rejected data was successful */
1873 if (unlikely(flags & MSG_PEEK))
1876 /* Send group flow control advertisement when applicable */
1877 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1878 __skb_queue_head_init(&xmitq);
1879 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1880 msg_orignode(hdr), msg_origport(hdr),
1882 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1885 tsk_advance_rx_queue(sk);
1887 if (likely(!connected))
1890 /* Send connection flow control advertisement when applicable */
1891 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1892 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1893 tipc_sk_send_ack(tsk);
1896 return rc ? rc : copy;
1900 * tipc_recvstream - receive stream-oriented data
1901 * @m: descriptor for message info
1902 * @buflen: total size of user buffer area
1903 * @flags: receive flags
1905 * Used for SOCK_STREAM messages only. If not enough data is available
1906 * will optionally wait for more; never truncates data.
1908 * Returns size of returned message data, errno otherwise
1910 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1911 size_t buflen, int flags)
1913 struct sock *sk = sock->sk;
1914 struct tipc_sock *tsk = tipc_sk(sk);
1915 struct sk_buff *skb;
1916 struct tipc_msg *hdr;
1917 struct tipc_skb_cb *skb_cb;
1918 bool peek = flags & MSG_PEEK;
1919 int offset, required, copy, copied = 0;
1920 int hlen, dlen, err, rc;
1924 /* Catch invalid receive attempts */
1925 if (unlikely(!buflen))
1930 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1934 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1935 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1938 /* Look at first msg in receive queue; wait if necessary */
1939 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1942 skb = skb_peek(&sk->sk_receive_queue);
1943 skb_cb = TIPC_SKB_CB(skb);
1945 dlen = msg_data_sz(hdr);
1946 hlen = msg_hdr_sz(hdr);
1947 err = msg_errcode(hdr);
1949 /* Discard any empty non-errored (SYN-) message */
1950 if (unlikely(!dlen && !err)) {
1951 tsk_advance_rx_queue(sk);
1955 /* Collect msg meta data, incl. error code and rejected data */
1957 tipc_sk_set_orig_addr(m, skb);
1958 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1964 /* Copy data if msg ok, otherwise return error/partial data */
1966 ack = msg_ack_required(hdr);
1967 offset = skb_cb->bytes_read;
1968 copy = min_t(int, dlen - offset, buflen - copied);
1969 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1974 if (unlikely(offset < dlen)) {
1976 skb_cb->bytes_read = offset;
1981 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1990 tsk_advance_rx_queue(sk);
1992 /* Send connection flow control advertisement when applicable */
1993 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1994 if (ack || tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1995 tipc_sk_send_ack(tsk);
1997 /* Exit if all requested data or FIN/error received */
1998 if (copied == buflen || err)
2001 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2004 return copied ? copied : rc;
2008 * tipc_write_space - wake up thread if port congestion is released
2011 static void tipc_write_space(struct sock *sk)
2013 struct socket_wq *wq;
2016 wq = rcu_dereference(sk->sk_wq);
2017 if (skwq_has_sleeper(wq))
2018 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2019 EPOLLWRNORM | EPOLLWRBAND);
2024 * tipc_data_ready - wake up threads to indicate messages have been received
2026 * @len: the length of messages
2028 static void tipc_data_ready(struct sock *sk)
2030 struct socket_wq *wq;
2033 wq = rcu_dereference(sk->sk_wq);
2034 if (skwq_has_sleeper(wq))
2035 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2036 EPOLLRDNORM | EPOLLRDBAND);
2040 static void tipc_sock_destruct(struct sock *sk)
2042 __skb_queue_purge(&sk->sk_receive_queue);
2045 static void tipc_sk_proto_rcv(struct sock *sk,
2046 struct sk_buff_head *inputq,
2047 struct sk_buff_head *xmitq)
2049 struct sk_buff *skb = __skb_dequeue(inputq);
2050 struct tipc_sock *tsk = tipc_sk(sk);
2051 struct tipc_msg *hdr = buf_msg(skb);
2052 struct tipc_group *grp = tsk->group;
2053 bool wakeup = false;
2055 switch (msg_user(hdr)) {
2057 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2060 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2061 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2063 tsk->cong_link_cnt--;
2065 tipc_sk_push_backlog(tsk);
2067 case GROUP_PROTOCOL:
2068 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2071 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2072 hdr, inputq, xmitq);
2079 sk->sk_write_space(sk);
2085 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2087 * @skb: pointer to message buffer.
2088 * Returns true if message should be added to receive queue, false otherwise
2090 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2092 struct sock *sk = &tsk->sk;
2093 struct net *net = sock_net(sk);
2094 struct tipc_msg *hdr = buf_msg(skb);
2095 bool con_msg = msg_connected(hdr);
2096 u32 pport = tsk_peer_port(tsk);
2097 u32 pnode = tsk_peer_node(tsk);
2098 u32 oport = msg_origport(hdr);
2099 u32 onode = msg_orignode(hdr);
2100 int err = msg_errcode(hdr);
2101 unsigned long delay;
2103 if (unlikely(msg_mcast(hdr)))
2107 switch (sk->sk_state) {
2108 case TIPC_CONNECTING:
2110 if (likely(con_msg)) {
2113 tipc_sk_finish_conn(tsk, oport, onode);
2114 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2115 /* ACK+ message with data is added to receive queue */
2116 if (msg_data_sz(hdr))
2118 /* Empty ACK-, - wake up sleeping connect() and drop */
2119 sk->sk_state_change(sk);
2120 msg_set_dest_droppable(hdr, 1);
2123 /* Ignore connectionless message if not from listening socket */
2124 if (oport != pport || onode != pnode)
2128 if (err != TIPC_ERR_OVERLOAD)
2131 /* Prepare for new setup attempt if we have a SYN clone */
2132 if (skb_queue_empty(&sk->sk_write_queue))
2134 get_random_bytes(&delay, 2);
2135 delay %= (tsk->conn_timeout / 4);
2136 delay = msecs_to_jiffies(delay + 100);
2137 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2140 case TIPC_DISCONNECTING:
2143 /* Accept only SYN message */
2144 if (!msg_is_syn(hdr) &&
2145 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2147 if (!con_msg && !err)
2150 case TIPC_ESTABLISHED:
2151 if (!skb_queue_empty(&sk->sk_write_queue))
2152 tipc_sk_push_backlog(tsk);
2153 /* Accept only connection-based messages sent by peer */
2154 if (likely(con_msg && !err && pport == oport && pnode == onode))
2156 if (!tsk_peer_msg(tsk, hdr))
2160 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2161 tipc_node_remove_conn(net, pnode, tsk->portid);
2162 sk->sk_state_change(sk);
2165 pr_err("Unknown sk_state %u\n", sk->sk_state);
2167 /* Abort connection setup attempt */
2168 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2169 sk->sk_err = ECONNREFUSED;
2170 sk->sk_state_change(sk);
2175 * rcvbuf_limit - get proper overload limit of socket receive queue
2179 * For connection oriented messages, irrespective of importance,
2180 * default queue limit is 2 MB.
2182 * For connectionless messages, queue limits are based on message
2183 * importance as follows:
2185 * TIPC_LOW_IMPORTANCE (2 MB)
2186 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2187 * TIPC_HIGH_IMPORTANCE (8 MB)
2188 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2190 * Returns overload limit according to corresponding message importance
2192 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2194 struct tipc_sock *tsk = tipc_sk(sk);
2195 struct tipc_msg *hdr = buf_msg(skb);
2197 if (unlikely(msg_in_group(hdr)))
2198 return READ_ONCE(sk->sk_rcvbuf);
2200 if (unlikely(!msg_connected(hdr)))
2201 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2203 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2204 return READ_ONCE(sk->sk_rcvbuf);
2206 return FLOWCTL_MSG_LIM;
2210 * tipc_sk_filter_rcv - validate incoming message
2212 * @skb: pointer to message.
2214 * Enqueues message on receive queue if acceptable; optionally handles
2215 * disconnect indication for a connected socket.
2217 * Called with socket lock already taken
2220 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2221 struct sk_buff_head *xmitq)
2223 bool sk_conn = !tipc_sk_type_connectionless(sk);
2224 struct tipc_sock *tsk = tipc_sk(sk);
2225 struct tipc_group *grp = tsk->group;
2226 struct tipc_msg *hdr = buf_msg(skb);
2227 struct net *net = sock_net(sk);
2228 struct sk_buff_head inputq;
2229 int mtyp = msg_type(hdr);
2230 int limit, err = TIPC_OK;
2232 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2233 TIPC_SKB_CB(skb)->bytes_read = 0;
2234 __skb_queue_head_init(&inputq);
2235 __skb_queue_tail(&inputq, skb);
2237 if (unlikely(!msg_isdata(hdr)))
2238 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2241 tipc_group_filter_msg(grp, &inputq, xmitq);
2243 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2244 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2246 /* Validate and add to receive buffer if there is space */
2247 while ((skb = __skb_dequeue(&inputq))) {
2249 limit = rcvbuf_limit(sk, skb);
2250 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2251 (!sk_conn && msg_connected(hdr)) ||
2252 (!grp && msg_in_group(hdr)))
2253 err = TIPC_ERR_NO_PORT;
2254 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2255 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2257 atomic_inc(&sk->sk_drops);
2258 err = TIPC_ERR_OVERLOAD;
2261 if (unlikely(err)) {
2262 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2263 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2265 __skb_queue_tail(xmitq, skb);
2270 __skb_queue_tail(&sk->sk_receive_queue, skb);
2271 skb_set_owner_r(skb, sk);
2272 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2273 "rcvq >90% allocated!");
2274 sk->sk_data_ready(sk);
2279 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2283 * Caller must hold socket lock
2285 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2287 unsigned int before = sk_rmem_alloc_get(sk);
2288 struct sk_buff_head xmitq;
2291 __skb_queue_head_init(&xmitq);
2293 tipc_sk_filter_rcv(sk, skb, &xmitq);
2294 added = sk_rmem_alloc_get(sk) - before;
2295 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2297 /* Send pending response/rejected messages, if any */
2298 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2303 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2304 * inputq and try adding them to socket or backlog queue
2305 * @inputq: list of incoming buffers with potentially different destinations
2306 * @sk: socket where the buffers should be enqueued
2307 * @dport: port number for the socket
2309 * Caller must hold socket lock
2311 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2312 u32 dport, struct sk_buff_head *xmitq)
2314 unsigned long time_limit = jiffies + 2;
2315 struct sk_buff *skb;
2320 while (skb_queue_len(inputq)) {
2321 if (unlikely(time_after_eq(jiffies, time_limit)))
2324 skb = tipc_skb_dequeue(inputq, dport);
2328 /* Add message directly to receive queue if possible */
2329 if (!sock_owned_by_user(sk)) {
2330 tipc_sk_filter_rcv(sk, skb, xmitq);
2334 /* Try backlog, compensating for double-counted bytes */
2335 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2336 if (!sk->sk_backlog.len)
2337 atomic_set(dcnt, 0);
2338 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2339 if (likely(!sk_add_backlog(sk, skb, lim))) {
2340 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2341 "bklg & rcvq >90% allocated!");
2345 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2346 /* Overload => reject message back to sender */
2347 onode = tipc_own_addr(sock_net(sk));
2348 atomic_inc(&sk->sk_drops);
2349 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2350 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2352 __skb_queue_tail(xmitq, skb);
2359 * tipc_sk_rcv - handle a chain of incoming buffers
2360 * @inputq: buffer list containing the buffers
2361 * Consumes all buffers in list until inputq is empty
2362 * Note: may be called in multiple threads referring to the same queue
2364 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2366 struct sk_buff_head xmitq;
2367 u32 dnode, dport = 0;
2369 struct tipc_sock *tsk;
2371 struct sk_buff *skb;
2373 __skb_queue_head_init(&xmitq);
2374 while (skb_queue_len(inputq)) {
2375 dport = tipc_skb_peek_port(inputq, dport);
2376 tsk = tipc_sk_lookup(net, dport);
2380 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2381 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2382 spin_unlock_bh(&sk->sk_lock.slock);
2384 /* Send pending response/rejected messages, if any */
2385 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2389 /* No destination socket => dequeue skb if still there */
2390 skb = tipc_skb_dequeue(inputq, dport);
2394 /* Try secondary lookup if unresolved named message */
2395 err = TIPC_ERR_NO_PORT;
2396 if (tipc_msg_lookup_dest(net, skb, &err))
2399 /* Prepare for message rejection */
2400 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2403 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2405 dnode = msg_destnode(buf_msg(skb));
2406 tipc_node_xmit_skb(net, skb, dnode, dport);
2410 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2412 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2413 struct sock *sk = sock->sk;
2417 int err = sock_error(sk);
2422 if (signal_pending(current))
2423 return sock_intr_errno(*timeo_p);
2425 add_wait_queue(sk_sleep(sk), &wait);
2426 done = sk_wait_event(sk, timeo_p,
2427 sk->sk_state != TIPC_CONNECTING, &wait);
2428 remove_wait_queue(sk_sleep(sk), &wait);
2433 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2435 if (addr->family != AF_TIPC)
2437 if (addr->addrtype == TIPC_SERVICE_RANGE)
2438 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2439 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2440 addr->addrtype == TIPC_SOCKET_ADDR);
2444 * tipc_connect - establish a connection to another TIPC port
2445 * @sock: socket structure
2446 * @dest: socket address for destination port
2447 * @destlen: size of socket address data structure
2448 * @flags: file-related flags associated with socket
2450 * Returns 0 on success, errno otherwise
2452 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2453 int destlen, int flags)
2455 struct sock *sk = sock->sk;
2456 struct tipc_sock *tsk = tipc_sk(sk);
2457 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2458 struct msghdr m = {NULL,};
2459 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2463 if (destlen != sizeof(struct sockaddr_tipc))
2473 if (dst->family == AF_UNSPEC) {
2474 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2475 if (!tipc_sk_type_connectionless(sk))
2479 if (!tipc_sockaddr_is_sane(dst)) {
2483 /* DGRAM/RDM connect(), just save the destaddr */
2484 if (tipc_sk_type_connectionless(sk)) {
2485 memcpy(&tsk->peer, dest, destlen);
2487 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2492 previous = sk->sk_state;
2494 switch (sk->sk_state) {
2496 /* Send a 'SYN-' to destination */
2498 m.msg_namelen = destlen;
2500 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2501 * indicate send_msg() is never blocked.
2504 m.msg_flags = MSG_DONTWAIT;
2506 res = __tipc_sendmsg(sock, &m, 0);
2507 if ((res < 0) && (res != -EWOULDBLOCK))
2510 /* Just entered TIPC_CONNECTING state; the only
2511 * difference is that return value in non-blocking
2512 * case is EINPROGRESS, rather than EALREADY.
2516 case TIPC_CONNECTING:
2518 if (previous == TIPC_CONNECTING)
2522 timeout = msecs_to_jiffies(timeout);
2523 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2524 res = tipc_wait_for_connect(sock, &timeout);
2526 case TIPC_ESTABLISHED:
2539 * tipc_listen - allow socket to listen for incoming connections
2540 * @sock: socket structure
2543 * Returns 0 on success, errno otherwise
2545 static int tipc_listen(struct socket *sock, int len)
2547 struct sock *sk = sock->sk;
2551 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2557 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2559 struct sock *sk = sock->sk;
2563 /* True wake-one mechanism for incoming connections: only
2564 * one process gets woken up, not the 'whole herd'.
2565 * Since we do not 'race & poll' for established sockets
2566 * anymore, the common case will execute the loop only once.
2569 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2570 TASK_INTERRUPTIBLE);
2571 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2573 timeo = schedule_timeout(timeo);
2577 if (!skb_queue_empty(&sk->sk_receive_queue))
2582 err = sock_intr_errno(timeo);
2583 if (signal_pending(current))
2586 finish_wait(sk_sleep(sk), &wait);
2591 * tipc_accept - wait for connection request
2592 * @sock: listening socket
2593 * @newsock: new socket that is to be connected
2594 * @flags: file-related flags associated with socket
2596 * Returns 0 on success, errno otherwise
2598 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2601 struct sock *new_sk, *sk = sock->sk;
2602 struct sk_buff *buf;
2603 struct tipc_sock *new_tsock;
2604 struct tipc_msg *msg;
2610 if (sk->sk_state != TIPC_LISTEN) {
2614 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2615 res = tipc_wait_for_accept(sock, timeo);
2619 buf = skb_peek(&sk->sk_receive_queue);
2621 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2624 security_sk_clone(sock->sk, new_sock->sk);
2626 new_sk = new_sock->sk;
2627 new_tsock = tipc_sk(new_sk);
2630 /* we lock on new_sk; but lockdep sees the lock on sk */
2631 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2634 * Reject any stray messages received by new socket
2635 * before the socket lock was taken (very, very unlikely)
2637 tsk_rej_rx_queue(new_sk);
2639 /* Connect new socket to it's peer */
2640 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2642 tsk_set_importance(new_tsock, msg_importance(msg));
2643 if (msg_named(msg)) {
2644 new_tsock->conn_type = msg_nametype(msg);
2645 new_tsock->conn_instance = msg_nameinst(msg);
2649 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2650 * Respond to 'SYN+' by queuing it on new socket.
2652 if (!msg_data_sz(msg)) {
2653 struct msghdr m = {NULL,};
2655 tsk_advance_rx_queue(sk);
2656 __tipc_sendstream(new_sock, &m, 0);
2658 __skb_dequeue(&sk->sk_receive_queue);
2659 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2660 skb_set_owner_r(buf, new_sk);
2662 release_sock(new_sk);
2669 * tipc_shutdown - shutdown socket connection
2670 * @sock: socket structure
2671 * @how: direction to close (must be SHUT_RDWR)
2673 * Terminates connection (if necessary), then purges socket's receive queue.
2675 * Returns 0 on success, errno otherwise
2677 static int tipc_shutdown(struct socket *sock, int how)
2679 struct sock *sk = sock->sk;
2682 if (how != SHUT_RDWR)
2687 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2688 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2689 sk->sk_shutdown = SEND_SHUTDOWN;
2691 if (sk->sk_state == TIPC_DISCONNECTING) {
2692 /* Discard any unreceived messages */
2693 __skb_queue_purge(&sk->sk_receive_queue);
2695 /* Wake up anyone sleeping in poll */
2696 sk->sk_state_change(sk);
2706 static void tipc_sk_check_probing_state(struct sock *sk,
2707 struct sk_buff_head *list)
2709 struct tipc_sock *tsk = tipc_sk(sk);
2710 u32 pnode = tsk_peer_node(tsk);
2711 u32 pport = tsk_peer_port(tsk);
2712 u32 self = tsk_own_node(tsk);
2713 u32 oport = tsk->portid;
2714 struct sk_buff *skb;
2716 if (tsk->probe_unacked) {
2717 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2718 sk->sk_err = ECONNABORTED;
2719 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2720 sk->sk_state_change(sk);
2723 /* Prepare new probe */
2724 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2725 pnode, self, pport, oport, TIPC_OK);
2727 __skb_queue_tail(list, skb);
2728 tsk->probe_unacked = true;
2729 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2732 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2734 struct tipc_sock *tsk = tipc_sk(sk);
2736 /* Try again later if dest link is congested */
2737 if (tsk->cong_link_cnt) {
2738 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2741 /* Prepare SYN for retransmit */
2742 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2745 static void tipc_sk_timeout(struct timer_list *t)
2747 struct sock *sk = from_timer(sk, t, sk_timer);
2748 struct tipc_sock *tsk = tipc_sk(sk);
2749 u32 pnode = tsk_peer_node(tsk);
2750 struct sk_buff_head list;
2753 __skb_queue_head_init(&list);
2756 /* Try again later if socket is busy */
2757 if (sock_owned_by_user(sk)) {
2758 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2763 if (sk->sk_state == TIPC_ESTABLISHED)
2764 tipc_sk_check_probing_state(sk, &list);
2765 else if (sk->sk_state == TIPC_CONNECTING)
2766 tipc_sk_retry_connect(sk, &list);
2770 if (!skb_queue_empty(&list))
2771 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2773 /* SYN messages may cause link congestion */
2774 if (rc == -ELINKCONG) {
2775 tipc_dest_push(&tsk->cong_links, pnode, 0);
2776 tsk->cong_link_cnt = 1;
2781 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2782 struct tipc_name_seq const *seq)
2784 struct sock *sk = &tsk->sk;
2785 struct net *net = sock_net(sk);
2786 struct publication *publ;
2789 if (scope != TIPC_NODE_SCOPE)
2790 scope = TIPC_CLUSTER_SCOPE;
2792 if (tipc_sk_connected(sk))
2794 key = tsk->portid + tsk->pub_count + 1;
2795 if (key == tsk->portid)
2798 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2799 scope, tsk->portid, key);
2800 if (unlikely(!publ))
2803 list_add(&publ->binding_sock, &tsk->publications);
2809 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2810 struct tipc_name_seq const *seq)
2812 struct net *net = sock_net(&tsk->sk);
2813 struct publication *publ;
2814 struct publication *safe;
2817 if (scope != TIPC_NODE_SCOPE)
2818 scope = TIPC_CLUSTER_SCOPE;
2820 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2822 if (publ->scope != scope)
2824 if (publ->type != seq->type)
2826 if (publ->lower != seq->lower)
2828 if (publ->upper != seq->upper)
2830 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2831 publ->upper, publ->key);
2835 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2836 publ->upper, publ->key);
2839 if (list_empty(&tsk->publications))
2844 /* tipc_sk_reinit: set non-zero address in all existing sockets
2845 * when we go from standalone to network mode.
2847 void tipc_sk_reinit(struct net *net)
2849 struct tipc_net *tn = net_generic(net, tipc_net_id);
2850 struct rhashtable_iter iter;
2851 struct tipc_sock *tsk;
2852 struct tipc_msg *msg;
2854 rhashtable_walk_enter(&tn->sk_rht, &iter);
2857 rhashtable_walk_start(&iter);
2859 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2860 sock_hold(&tsk->sk);
2861 rhashtable_walk_stop(&iter);
2862 lock_sock(&tsk->sk);
2864 msg_set_prevnode(msg, tipc_own_addr(net));
2865 msg_set_orignode(msg, tipc_own_addr(net));
2866 release_sock(&tsk->sk);
2867 rhashtable_walk_start(&iter);
2871 rhashtable_walk_stop(&iter);
2872 } while (tsk == ERR_PTR(-EAGAIN));
2874 rhashtable_walk_exit(&iter);
2877 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2879 struct tipc_net *tn = net_generic(net, tipc_net_id);
2880 struct tipc_sock *tsk;
2883 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2885 sock_hold(&tsk->sk);
2891 static int tipc_sk_insert(struct tipc_sock *tsk)
2893 struct sock *sk = &tsk->sk;
2894 struct net *net = sock_net(sk);
2895 struct tipc_net *tn = net_generic(net, tipc_net_id);
2896 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2897 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2899 while (remaining--) {
2901 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2902 portid = TIPC_MIN_PORT;
2903 tsk->portid = portid;
2904 sock_hold(&tsk->sk);
2905 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2914 static void tipc_sk_remove(struct tipc_sock *tsk)
2916 struct sock *sk = &tsk->sk;
2917 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2919 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2920 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2925 static const struct rhashtable_params tsk_rht_params = {
2927 .head_offset = offsetof(struct tipc_sock, node),
2928 .key_offset = offsetof(struct tipc_sock, portid),
2929 .key_len = sizeof(u32), /* portid */
2930 .max_size = 1048576,
2932 .automatic_shrinking = true,
2935 int tipc_sk_rht_init(struct net *net)
2937 struct tipc_net *tn = net_generic(net, tipc_net_id);
2939 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2942 void tipc_sk_rht_destroy(struct net *net)
2944 struct tipc_net *tn = net_generic(net, tipc_net_id);
2946 /* Wait for socket readers to complete */
2949 rhashtable_destroy(&tn->sk_rht);
2952 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2954 struct net *net = sock_net(&tsk->sk);
2955 struct tipc_group *grp = tsk->group;
2956 struct tipc_msg *hdr = &tsk->phdr;
2957 struct tipc_name_seq seq;
2960 if (mreq->type < TIPC_RESERVED_TYPES)
2962 if (mreq->scope > TIPC_NODE_SCOPE)
2966 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2970 msg_set_lookup_scope(hdr, mreq->scope);
2971 msg_set_nametype(hdr, mreq->type);
2972 msg_set_dest_droppable(hdr, true);
2973 seq.type = mreq->type;
2974 seq.lower = mreq->instance;
2975 seq.upper = seq.lower;
2976 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2977 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2979 tipc_group_delete(net, grp);
2983 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2984 tsk->mc_method.rcast = true;
2985 tsk->mc_method.mandatory = true;
2986 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2990 static int tipc_sk_leave(struct tipc_sock *tsk)
2992 struct net *net = sock_net(&tsk->sk);
2993 struct tipc_group *grp = tsk->group;
2994 struct tipc_name_seq seq;
2999 tipc_group_self(grp, &seq, &scope);
3000 tipc_group_delete(net, grp);
3002 tipc_sk_withdraw(tsk, scope, &seq);
3007 * tipc_setsockopt - set socket option
3008 * @sock: socket structure
3009 * @lvl: option level
3010 * @opt: option identifier
3011 * @ov: pointer to new option value
3012 * @ol: length of option value
3014 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3015 * (to ease compatibility).
3017 * Returns 0 on success, errno otherwise
3019 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3020 char __user *ov, unsigned int ol)
3022 struct sock *sk = sock->sk;
3023 struct tipc_sock *tsk = tipc_sk(sk);
3024 struct tipc_group_req mreq;
3028 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3030 if (lvl != SOL_TIPC)
3031 return -ENOPROTOOPT;
3034 case TIPC_IMPORTANCE:
3035 case TIPC_SRC_DROPPABLE:
3036 case TIPC_DEST_DROPPABLE:
3037 case TIPC_CONN_TIMEOUT:
3039 if (ol < sizeof(value))
3041 if (get_user(value, (u32 __user *)ov))
3044 case TIPC_GROUP_JOIN:
3045 if (ol < sizeof(mreq))
3047 if (copy_from_user(&mreq, ov, sizeof(mreq)))
3058 case TIPC_IMPORTANCE:
3059 res = tsk_set_importance(tsk, value);
3061 case TIPC_SRC_DROPPABLE:
3062 if (sock->type != SOCK_STREAM)
3063 tsk_set_unreliable(tsk, value);
3067 case TIPC_DEST_DROPPABLE:
3068 tsk_set_unreturnable(tsk, value);
3070 case TIPC_CONN_TIMEOUT:
3071 tipc_sk(sk)->conn_timeout = value;
3073 case TIPC_MCAST_BROADCAST:
3074 tsk->mc_method.rcast = false;
3075 tsk->mc_method.mandatory = true;
3077 case TIPC_MCAST_REPLICAST:
3078 tsk->mc_method.rcast = true;
3079 tsk->mc_method.mandatory = true;
3081 case TIPC_GROUP_JOIN:
3082 res = tipc_sk_join(tsk, &mreq);
3084 case TIPC_GROUP_LEAVE:
3085 res = tipc_sk_leave(tsk);
3088 tsk->nodelay = !!value;
3101 * tipc_getsockopt - get socket option
3102 * @sock: socket structure
3103 * @lvl: option level
3104 * @opt: option identifier
3105 * @ov: receptacle for option value
3106 * @ol: receptacle for length of option value
3108 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3109 * (to ease compatibility).
3111 * Returns 0 on success, errno otherwise
3113 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3114 char __user *ov, int __user *ol)
3116 struct sock *sk = sock->sk;
3117 struct tipc_sock *tsk = tipc_sk(sk);
3118 struct tipc_name_seq seq;
3123 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3124 return put_user(0, ol);
3125 if (lvl != SOL_TIPC)
3126 return -ENOPROTOOPT;
3127 res = get_user(len, ol);
3134 case TIPC_IMPORTANCE:
3135 value = tsk_importance(tsk);
3137 case TIPC_SRC_DROPPABLE:
3138 value = tsk_unreliable(tsk);
3140 case TIPC_DEST_DROPPABLE:
3141 value = tsk_unreturnable(tsk);
3143 case TIPC_CONN_TIMEOUT:
3144 value = tsk->conn_timeout;
3145 /* no need to set "res", since already 0 at this point */
3147 case TIPC_NODE_RECVQ_DEPTH:
3148 value = 0; /* was tipc_queue_size, now obsolete */
3150 case TIPC_SOCK_RECVQ_DEPTH:
3151 value = skb_queue_len(&sk->sk_receive_queue);
3153 case TIPC_SOCK_RECVQ_USED:
3154 value = sk_rmem_alloc_get(sk);
3156 case TIPC_GROUP_JOIN:
3159 tipc_group_self(tsk->group, &seq, &scope);
3169 return res; /* "get" failed */
3171 if (len < sizeof(value))
3174 if (copy_to_user(ov, &value, sizeof(value)))
3177 return put_user(sizeof(value), ol);
3180 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3182 struct net *net = sock_net(sock->sk);
3183 struct tipc_sioc_nodeid_req nr = {0};
3184 struct tipc_sioc_ln_req lnr;
3185 void __user *argp = (void __user *)arg;
3188 case SIOCGETLINKNAME:
3189 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3191 if (!tipc_node_get_linkname(net,
3192 lnr.bearer_id & 0xffff, lnr.peer,
3193 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3194 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3198 return -EADDRNOTAVAIL;
3200 if (copy_from_user(&nr, argp, sizeof(nr)))
3202 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3203 return -EADDRNOTAVAIL;
3204 if (copy_to_user(argp, &nr, sizeof(nr)))
3208 return -ENOIOCTLCMD;
3212 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3214 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3215 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3216 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3218 tsk1->peer.family = AF_TIPC;
3219 tsk1->peer.addrtype = TIPC_ADDR_ID;
3220 tsk1->peer.scope = TIPC_NODE_SCOPE;
3221 tsk1->peer.addr.id.ref = tsk2->portid;
3222 tsk1->peer.addr.id.node = onode;
3223 tsk2->peer.family = AF_TIPC;
3224 tsk2->peer.addrtype = TIPC_ADDR_ID;
3225 tsk2->peer.scope = TIPC_NODE_SCOPE;
3226 tsk2->peer.addr.id.ref = tsk1->portid;
3227 tsk2->peer.addr.id.node = onode;
3229 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3230 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3234 /* Protocol switches for the various types of TIPC sockets */
3236 static const struct proto_ops msg_ops = {
3237 .owner = THIS_MODULE,
3239 .release = tipc_release,
3241 .connect = tipc_connect,
3242 .socketpair = tipc_socketpair,
3243 .accept = sock_no_accept,
3244 .getname = tipc_getname,
3246 .ioctl = tipc_ioctl,
3247 .listen = sock_no_listen,
3248 .shutdown = tipc_shutdown,
3249 .setsockopt = tipc_setsockopt,
3250 .getsockopt = tipc_getsockopt,
3251 .sendmsg = tipc_sendmsg,
3252 .recvmsg = tipc_recvmsg,
3253 .mmap = sock_no_mmap,
3254 .sendpage = sock_no_sendpage
3257 static const struct proto_ops packet_ops = {
3258 .owner = THIS_MODULE,
3260 .release = tipc_release,
3262 .connect = tipc_connect,
3263 .socketpair = tipc_socketpair,
3264 .accept = tipc_accept,
3265 .getname = tipc_getname,
3267 .ioctl = tipc_ioctl,
3268 .listen = tipc_listen,
3269 .shutdown = tipc_shutdown,
3270 .setsockopt = tipc_setsockopt,
3271 .getsockopt = tipc_getsockopt,
3272 .sendmsg = tipc_send_packet,
3273 .recvmsg = tipc_recvmsg,
3274 .mmap = sock_no_mmap,
3275 .sendpage = sock_no_sendpage
3278 static const struct proto_ops stream_ops = {
3279 .owner = THIS_MODULE,
3281 .release = tipc_release,
3283 .connect = tipc_connect,
3284 .socketpair = tipc_socketpair,
3285 .accept = tipc_accept,
3286 .getname = tipc_getname,
3288 .ioctl = tipc_ioctl,
3289 .listen = tipc_listen,
3290 .shutdown = tipc_shutdown,
3291 .setsockopt = tipc_setsockopt,
3292 .getsockopt = tipc_getsockopt,
3293 .sendmsg = tipc_sendstream,
3294 .recvmsg = tipc_recvstream,
3295 .mmap = sock_no_mmap,
3296 .sendpage = sock_no_sendpage
3299 static const struct net_proto_family tipc_family_ops = {
3300 .owner = THIS_MODULE,
3302 .create = tipc_sk_create
3305 static struct proto tipc_proto = {
3307 .owner = THIS_MODULE,
3308 .obj_size = sizeof(struct tipc_sock),
3309 .sysctl_rmem = sysctl_tipc_rmem
3313 * tipc_socket_init - initialize TIPC socket interface
3315 * Returns 0 on success, errno otherwise
3317 int tipc_socket_init(void)
3321 res = proto_register(&tipc_proto, 1);
3323 pr_err("Failed to register TIPC protocol type\n");
3327 res = sock_register(&tipc_family_ops);
3329 pr_err("Failed to register TIPC socket type\n");
3330 proto_unregister(&tipc_proto);
3338 * tipc_socket_stop - stop TIPC socket interface
3340 void tipc_socket_stop(void)
3342 sock_unregister(tipc_family_ops.family);
3343 proto_unregister(&tipc_proto);
3346 /* Caller should hold socket lock for the passed tipc socket. */
3347 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3351 struct nlattr *nest;
3353 peer_node = tsk_peer_node(tsk);
3354 peer_port = tsk_peer_port(tsk);
3356 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3360 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3362 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3365 if (tsk->conn_type != 0) {
3366 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3368 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3370 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3373 nla_nest_end(skb, nest);
3378 nla_nest_cancel(skb, nest);
3383 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3386 struct net *net = sock_net(skb->sk);
3387 struct sock *sk = &tsk->sk;
3389 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3390 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3393 if (tipc_sk_connected(sk)) {
3394 if (__tipc_nl_add_sk_con(skb, tsk))
3396 } else if (!list_empty(&tsk->publications)) {
3397 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3403 /* Caller should hold socket lock for the passed tipc socket. */
3404 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3405 struct tipc_sock *tsk)
3407 struct nlattr *attrs;
3410 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3411 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3415 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3417 goto genlmsg_cancel;
3419 if (__tipc_nl_add_sk_info(skb, tsk))
3420 goto attr_msg_cancel;
3422 nla_nest_end(skb, attrs);
3423 genlmsg_end(skb, hdr);
3428 nla_nest_cancel(skb, attrs);
3430 genlmsg_cancel(skb, hdr);
3435 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3436 int (*skb_handler)(struct sk_buff *skb,
3437 struct netlink_callback *cb,
3438 struct tipc_sock *tsk))
3440 struct rhashtable_iter *iter = (void *)cb->args[4];
3441 struct tipc_sock *tsk;
3444 rhashtable_walk_start(iter);
3445 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3448 if (err == -EAGAIN) {
3455 sock_hold(&tsk->sk);
3456 rhashtable_walk_stop(iter);
3457 lock_sock(&tsk->sk);
3458 err = skb_handler(skb, cb, tsk);
3460 release_sock(&tsk->sk);
3464 release_sock(&tsk->sk);
3465 rhashtable_walk_start(iter);
3468 rhashtable_walk_stop(iter);
3472 EXPORT_SYMBOL(tipc_nl_sk_walk);
3474 int tipc_dump_start(struct netlink_callback *cb)
3476 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3478 EXPORT_SYMBOL(tipc_dump_start);
3480 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3482 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3483 struct rhashtable_iter *iter = (void *)cb->args[4];
3484 struct tipc_net *tn = tipc_net(net);
3487 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3491 cb->args[4] = (long)iter;
3494 rhashtable_walk_enter(&tn->sk_rht, iter);
3498 int tipc_dump_done(struct netlink_callback *cb)
3500 struct rhashtable_iter *hti = (void *)cb->args[4];
3502 rhashtable_walk_exit(hti);
3506 EXPORT_SYMBOL(tipc_dump_done);
3508 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3509 struct tipc_sock *tsk, u32 sk_filter_state,
3510 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3512 struct sock *sk = &tsk->sk;
3513 struct nlattr *attrs;
3514 struct nlattr *stat;
3516 /*filter response w.r.t sk_state*/
3517 if (!(sk_filter_state & (1 << sk->sk_state)))
3520 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3524 if (__tipc_nl_add_sk_info(skb, tsk))
3525 goto attr_msg_cancel;
3527 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3528 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3529 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3530 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3531 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3533 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3534 tipc_diag_gen_cookie(sk),
3536 goto attr_msg_cancel;
3538 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3540 goto attr_msg_cancel;
3542 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3543 skb_queue_len(&sk->sk_receive_queue)) ||
3544 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3545 skb_queue_len(&sk->sk_write_queue)) ||
3546 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3547 atomic_read(&sk->sk_drops)))
3548 goto stat_msg_cancel;
3550 if (tsk->cong_link_cnt &&
3551 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3552 goto stat_msg_cancel;
3554 if (tsk_conn_cong(tsk) &&
3555 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3556 goto stat_msg_cancel;
3558 nla_nest_end(skb, stat);
3561 if (tipc_group_fill_sock_diag(tsk->group, skb))
3562 goto stat_msg_cancel;
3564 nla_nest_end(skb, attrs);
3569 nla_nest_cancel(skb, stat);
3571 nla_nest_cancel(skb, attrs);
3575 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3577 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3579 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3582 /* Caller should hold socket lock for the passed tipc socket. */
3583 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3584 struct netlink_callback *cb,
3585 struct publication *publ)
3588 struct nlattr *attrs;
3590 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3591 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3595 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3597 goto genlmsg_cancel;
3599 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3600 goto attr_msg_cancel;
3601 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3602 goto attr_msg_cancel;
3603 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3604 goto attr_msg_cancel;
3605 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3606 goto attr_msg_cancel;
3608 nla_nest_end(skb, attrs);
3609 genlmsg_end(skb, hdr);
3614 nla_nest_cancel(skb, attrs);
3616 genlmsg_cancel(skb, hdr);
3621 /* Caller should hold socket lock for the passed tipc socket. */
3622 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3623 struct netlink_callback *cb,
3624 struct tipc_sock *tsk, u32 *last_publ)
3627 struct publication *p;
3630 list_for_each_entry(p, &tsk->publications, binding_sock) {
3631 if (p->key == *last_publ)
3634 if (p->key != *last_publ) {
3635 /* We never set seq or call nl_dump_check_consistent()
3636 * this means that setting prev_seq here will cause the
3637 * consistence check to fail in the netlink callback
3638 * handler. Resulting in the last NLMSG_DONE message
3639 * having the NLM_F_DUMP_INTR flag set.
3646 p = list_first_entry(&tsk->publications, struct publication,
3650 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3651 err = __tipc_nl_add_sk_publ(skb, cb, p);
3653 *last_publ = p->key;
3662 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3665 u32 tsk_portid = cb->args[0];
3666 u32 last_publ = cb->args[1];
3667 u32 done = cb->args[2];
3668 struct net *net = sock_net(skb->sk);
3669 struct tipc_sock *tsk;
3672 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3673 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3675 if (!attrs[TIPC_NLA_SOCK])
3678 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3679 attrs[TIPC_NLA_SOCK],
3680 tipc_nl_sock_policy, NULL);
3684 if (!sock[TIPC_NLA_SOCK_REF])
3687 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3693 tsk = tipc_sk_lookup(net, tsk_portid);
3697 lock_sock(&tsk->sk);
3698 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3701 release_sock(&tsk->sk);
3704 cb->args[0] = tsk_portid;
3705 cb->args[1] = last_publ;
3712 * tipc_sk_filtering - check if a socket should be traced
3713 * @sk: the socket to be examined
3714 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3715 * (portid, sock type, name type, name lower, name upper)
3717 * Returns true if the socket meets the socket tuple data
3718 * (value 0 = 'any') or when there is no tuple set (all = 0),
3721 bool tipc_sk_filtering(struct sock *sk)
3723 struct tipc_sock *tsk;
3724 struct publication *p;
3725 u32 _port, _sktype, _type, _lower, _upper;
3726 u32 type = 0, lower = 0, upper = 0;
3733 _port = sysctl_tipc_sk_filter[0];
3734 _sktype = sysctl_tipc_sk_filter[1];
3735 _type = sysctl_tipc_sk_filter[2];
3736 _lower = sysctl_tipc_sk_filter[3];
3737 _upper = sysctl_tipc_sk_filter[4];
3739 if (!_port && !_sktype && !_type && !_lower && !_upper)
3743 return (_port == tsk->portid);
3745 if (_sktype && _sktype != sk->sk_type)
3748 if (tsk->published) {
3749 p = list_first_entry_or_null(&tsk->publications,
3750 struct publication, binding_sock);
3758 if (!tipc_sk_type_connectionless(sk)) {
3759 type = tsk->conn_type;
3760 lower = tsk->conn_instance;
3761 upper = tsk->conn_instance;
3764 if ((_type && _type != type) || (_lower && _lower != lower) ||
3765 (_upper && _upper != upper))
3771 u32 tipc_sock_get_portid(struct sock *sk)
3773 return (sk) ? (tipc_sk(sk))->portid : 0;
3777 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3778 * both the rcv and backlog queues are considered
3779 * @sk: tipc sk to be checked
3780 * @skb: tipc msg to be checked
3782 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3785 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3787 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3788 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3789 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3791 return (qsize > lim * 90 / 100);
3795 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3796 * only the rcv queue is considered
3797 * @sk: tipc sk to be checked
3798 * @skb: tipc msg to be checked
3800 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3803 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3805 unsigned int lim = rcvbuf_limit(sk, skb);
3806 unsigned int qsize = sk_rmem_alloc_get(sk);
3808 return (qsize > lim * 90 / 100);
3812 * tipc_sk_dump - dump TIPC socket
3813 * @sk: tipc sk to be dumped
3814 * @dqueues: bitmask to decide if any socket queue to be dumped?
3815 * - TIPC_DUMP_NONE: don't dump socket queues
3816 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3817 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3818 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3819 * - TIPC_DUMP_ALL: dump all the socket queues above
3820 * @buf: returned buffer of dump data in format
3822 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3825 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3826 struct tipc_sock *tsk;
3827 struct publication *p;
3831 i += scnprintf(buf, sz, "sk data: (null)\n");
3836 tsk_connected = !tipc_sk_type_connectionless(sk);
3838 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3839 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3840 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3841 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3842 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3843 if (tsk_connected) {
3844 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3845 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3846 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3847 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3849 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3850 if (tsk->published) {
3851 p = list_first_entry_or_null(&tsk->publications,
3852 struct publication, binding_sock);
3853 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3854 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3855 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3857 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3858 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3859 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3860 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3861 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3862 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3863 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3864 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3865 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3866 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3867 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3868 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3869 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3870 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3872 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3873 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3874 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3877 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3878 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3879 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3882 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3883 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3884 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3885 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3886 i += scnprintf(buf + i, sz - i, " tail ");
3887 i += tipc_skb_dump(sk->sk_backlog.tail, false,