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 NAGLE_START_INIT 4
52 #define NAGLE_START_MAX 1024
53 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
54 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
55 #define TIPC_FWD_MSG 1
56 #define TIPC_MAX_PORT 0xffffffff
57 #define TIPC_MIN_PORT 1
58 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
61 TIPC_LISTEN = TCP_LISTEN,
62 TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 TIPC_OPEN = TCP_CLOSE,
64 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 TIPC_CONNECTING = TCP_SYN_SENT,
68 struct sockaddr_pair {
69 struct sockaddr_tipc sock;
70 struct sockaddr_tipc member;
74 * struct tipc_sock - TIPC socket structure
75 * @sk: socket - interacts with 'port' and with user via the socket API
76 * @conn_type: TIPC type used when connection was established
77 * @conn_instance: TIPC instance used when connection was established
78 * @published: non-zero if port has one or more associated names
79 * @max_pkt: maximum packet size "hint" used when building messages sent by port
80 * @maxnagle: maximum size of msg which can be subject to nagle
81 * @portid: unique port identity in TIPC socket hash table
82 * @phdr: preformatted message header used when sending messages
83 * #cong_links: list of congested links
84 * @publications: list of publications for port
85 * @blocking_link: address of the congested link we are currently sleeping on
86 * @pub_count: total # of publications port has made during its lifetime
87 * @conn_timeout: the time we can wait for an unresponded setup request
88 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
89 * @cong_link_cnt: number of congested links
90 * @snt_unacked: # messages sent by socket, and not yet acked by peer
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @peer: 'connected' peer for dgram/rdm
93 * @node: hash table node
94 * @mc_method: cookie for use between socket and broadcast layer
95 * @rcu: rcu struct for tipc_sock
105 struct tipc_msg phdr;
106 struct list_head cong_links;
107 struct list_head publications;
109 atomic_t dupl_rcvcnt;
118 struct sockaddr_tipc peer;
119 struct rhash_head node;
120 struct tipc_mc_method mc_method;
122 struct tipc_group *group;
133 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
134 static void tipc_data_ready(struct sock *sk);
135 static void tipc_write_space(struct sock *sk);
136 static void tipc_sock_destruct(struct sock *sk);
137 static int tipc_release(struct socket *sock);
138 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
140 static void tipc_sk_timeout(struct timer_list *t);
141 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
142 struct tipc_name_seq const *seq);
143 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
144 struct tipc_name_seq const *seq);
145 static int tipc_sk_leave(struct tipc_sock *tsk);
146 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
147 static int tipc_sk_insert(struct tipc_sock *tsk);
148 static void tipc_sk_remove(struct tipc_sock *tsk);
149 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
150 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
151 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
153 static const struct proto_ops packet_ops;
154 static const struct proto_ops stream_ops;
155 static const struct proto_ops msg_ops;
156 static struct proto tipc_proto;
157 static const struct rhashtable_params tsk_rht_params;
159 static u32 tsk_own_node(struct tipc_sock *tsk)
161 return msg_prevnode(&tsk->phdr);
164 static u32 tsk_peer_node(struct tipc_sock *tsk)
166 return msg_destnode(&tsk->phdr);
169 static u32 tsk_peer_port(struct tipc_sock *tsk)
171 return msg_destport(&tsk->phdr);
174 static bool tsk_unreliable(struct tipc_sock *tsk)
176 return msg_src_droppable(&tsk->phdr) != 0;
179 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
181 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
184 static bool tsk_unreturnable(struct tipc_sock *tsk)
186 return msg_dest_droppable(&tsk->phdr) != 0;
189 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
191 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
194 static int tsk_importance(struct tipc_sock *tsk)
196 return msg_importance(&tsk->phdr);
199 static struct tipc_sock *tipc_sk(const struct sock *sk)
201 return container_of(sk, struct tipc_sock, sk);
204 int tsk_set_importance(struct sock *sk, int imp)
206 if (imp > TIPC_CRITICAL_IMPORTANCE)
208 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
212 static bool tsk_conn_cong(struct tipc_sock *tsk)
214 return tsk->snt_unacked > tsk->snd_win;
217 static u16 tsk_blocks(int len)
219 return ((len / FLOWCTL_BLK_SZ) + 1);
222 /* tsk_blocks(): translate a buffer size in bytes to number of
223 * advertisable blocks, taking into account the ratio truesize(len)/len
224 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
226 static u16 tsk_adv_blocks(int len)
228 return len / FLOWCTL_BLK_SZ / 4;
231 /* tsk_inc(): increment counter for sent or received data
232 * - If block based flow control is not supported by peer we
233 * fall back to message based ditto, incrementing the counter
235 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
237 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
238 return ((msglen / FLOWCTL_BLK_SZ) + 1);
242 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
244 static void tsk_set_nagle(struct tipc_sock *tsk)
246 struct sock *sk = &tsk->sk;
249 if (sk->sk_type != SOCK_STREAM)
253 if (!(tsk->peer_caps & TIPC_NAGLE))
255 /* Limit node local buffer size to avoid receive queue overflow */
256 if (tsk->max_pkt == MAX_MSG_SIZE)
257 tsk->maxnagle = 1500;
259 tsk->maxnagle = tsk->max_pkt;
263 * tsk_advance_rx_queue - discard first buffer in socket receive queue
265 * Caller must hold socket lock
267 static void tsk_advance_rx_queue(struct sock *sk)
269 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
270 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
273 /* tipc_sk_respond() : send response message back to sender
275 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
279 u32 onode = tipc_own_addr(sock_net(sk));
281 if (!tipc_msg_reverse(onode, &skb, err))
284 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
285 dnode = msg_destnode(buf_msg(skb));
286 selector = msg_origport(buf_msg(skb));
287 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
291 * tsk_rej_rx_queue - reject all buffers in socket receive queue
293 * Caller must hold socket lock
295 static void tsk_rej_rx_queue(struct sock *sk, int error)
299 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
300 tipc_sk_respond(sk, skb, error);
303 static bool tipc_sk_connected(struct sock *sk)
305 return sk->sk_state == TIPC_ESTABLISHED;
308 /* tipc_sk_type_connectionless - check if the socket is datagram socket
311 * Returns true if connection less, false otherwise
313 static bool tipc_sk_type_connectionless(struct sock *sk)
315 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
318 /* tsk_peer_msg - verify if message was sent by connected port's peer
320 * Handles cases where the node's network address has changed from
321 * the default of <0.0.0> to its configured setting.
323 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
325 struct sock *sk = &tsk->sk;
326 u32 self = tipc_own_addr(sock_net(sk));
327 u32 peer_port = tsk_peer_port(tsk);
328 u32 orig_node, peer_node;
330 if (unlikely(!tipc_sk_connected(sk)))
333 if (unlikely(msg_origport(msg) != peer_port))
336 orig_node = msg_orignode(msg);
337 peer_node = tsk_peer_node(tsk);
339 if (likely(orig_node == peer_node))
342 if (!orig_node && peer_node == self)
345 if (!peer_node && orig_node == self)
351 /* tipc_set_sk_state - set the sk_state of the socket
354 * Caller must hold socket lock
356 * Returns 0 on success, errno otherwise
358 static int tipc_set_sk_state(struct sock *sk, int state)
360 int oldsk_state = sk->sk_state;
368 case TIPC_CONNECTING:
369 if (oldsk_state == TIPC_OPEN)
372 case TIPC_ESTABLISHED:
373 if (oldsk_state == TIPC_CONNECTING ||
374 oldsk_state == TIPC_OPEN)
377 case TIPC_DISCONNECTING:
378 if (oldsk_state == TIPC_CONNECTING ||
379 oldsk_state == TIPC_ESTABLISHED)
385 sk->sk_state = state;
390 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
392 struct sock *sk = sock->sk;
393 int err = sock_error(sk);
394 int typ = sock->type;
398 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
399 if (sk->sk_state == TIPC_DISCONNECTING)
401 else if (!tipc_sk_connected(sk))
406 if (signal_pending(current))
407 return sock_intr_errno(*timeout);
412 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
414 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
418 while ((rc_ = !(condition_))) { \
419 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
422 rc_ = tipc_sk_sock_err((sock_), timeo_); \
425 add_wait_queue(sk_sleep(sk_), &wait_); \
427 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
428 sched_annotate_sleep(); \
430 remove_wait_queue(sk_sleep(sk_), &wait_); \
436 * tipc_sk_create - create a TIPC socket
437 * @net: network namespace (must be default network)
438 * @sock: pre-allocated socket structure
439 * @protocol: protocol indicator (must be 0)
440 * @kern: caused by kernel or by userspace?
442 * This routine creates additional data structures used by the TIPC socket,
443 * initializes them, and links them together.
445 * Returns 0 on success, errno otherwise
447 static int tipc_sk_create(struct net *net, struct socket *sock,
448 int protocol, int kern)
450 const struct proto_ops *ops;
452 struct tipc_sock *tsk;
453 struct tipc_msg *msg;
455 /* Validate arguments */
456 if (unlikely(protocol != 0))
457 return -EPROTONOSUPPORT;
459 switch (sock->type) {
474 /* Allocate socket's protocol area */
475 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
480 tsk->max_pkt = MAX_PKT_DEFAULT;
482 tsk->nagle_start = NAGLE_START_INIT;
483 INIT_LIST_HEAD(&tsk->publications);
484 INIT_LIST_HEAD(&tsk->cong_links);
487 /* Finish initializing socket data structures */
489 sock_init_data(sock, sk);
490 tipc_set_sk_state(sk, TIPC_OPEN);
491 if (tipc_sk_insert(tsk)) {
492 pr_warn("Socket create failed; port number exhausted\n");
496 /* Ensure tsk is visible before we read own_addr. */
499 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
500 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
502 msg_set_origport(msg, tsk->portid);
503 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
505 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
506 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
507 sk->sk_data_ready = tipc_data_ready;
508 sk->sk_write_space = tipc_write_space;
509 sk->sk_destruct = tipc_sock_destruct;
510 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
511 tsk->group_is_open = true;
512 atomic_set(&tsk->dupl_rcvcnt, 0);
514 /* Start out with safe limits until we receive an advertised window */
515 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
516 tsk->rcv_win = tsk->snd_win;
518 if (tipc_sk_type_connectionless(sk)) {
519 tsk_set_unreturnable(tsk, true);
520 if (sock->type == SOCK_DGRAM)
521 tsk_set_unreliable(tsk, true);
523 __skb_queue_head_init(&tsk->mc_method.deferredq);
524 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
528 static void tipc_sk_callback(struct rcu_head *head)
530 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
535 /* Caller should hold socket lock for the socket. */
536 static void __tipc_shutdown(struct socket *sock, int error)
538 struct sock *sk = sock->sk;
539 struct tipc_sock *tsk = tipc_sk(sk);
540 struct net *net = sock_net(sk);
541 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
542 u32 dnode = tsk_peer_node(tsk);
545 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
546 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
547 !tsk_conn_cong(tsk)));
549 /* Push out delayed messages if in Nagle mode */
550 tipc_sk_push_backlog(tsk, false);
551 /* Remove pending SYN */
552 __skb_queue_purge(&sk->sk_write_queue);
554 /* Remove partially received buffer if any */
555 skb = skb_peek(&sk->sk_receive_queue);
556 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
557 __skb_unlink(skb, &sk->sk_receive_queue);
561 /* Reject all unreceived messages if connectionless */
562 if (tipc_sk_type_connectionless(sk)) {
563 tsk_rej_rx_queue(sk, error);
567 switch (sk->sk_state) {
568 case TIPC_CONNECTING:
569 case TIPC_ESTABLISHED:
570 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
571 tipc_node_remove_conn(net, dnode, tsk->portid);
572 /* Send a FIN+/- to its peer */
573 skb = __skb_dequeue(&sk->sk_receive_queue);
575 __skb_queue_purge(&sk->sk_receive_queue);
576 tipc_sk_respond(sk, skb, error);
579 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
580 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
581 tsk_own_node(tsk), tsk_peer_port(tsk),
584 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
587 /* Reject all SYN messages */
588 tsk_rej_rx_queue(sk, error);
591 __skb_queue_purge(&sk->sk_receive_queue);
597 * tipc_release - destroy a TIPC socket
598 * @sock: socket to destroy
600 * This routine cleans up any messages that are still queued on the socket.
601 * For DGRAM and RDM socket types, all queued messages are rejected.
602 * For SEQPACKET and STREAM socket types, the first message is rejected
603 * and any others are discarded. (If the first message on a STREAM socket
604 * is partially-read, it is discarded and the next one is rejected instead.)
606 * NOTE: Rejected messages are not necessarily returned to the sender! They
607 * are returned or discarded according to the "destination droppable" setting
608 * specified for the message by the sender.
610 * Returns 0 on success, errno otherwise
612 static int tipc_release(struct socket *sock)
614 struct sock *sk = sock->sk;
615 struct tipc_sock *tsk;
618 * Exit if socket isn't fully initialized (occurs when a failed accept()
619 * releases a pre-allocated child socket that was never used)
627 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
628 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
629 sk->sk_shutdown = SHUTDOWN_MASK;
631 tipc_sk_withdraw(tsk, 0, NULL);
632 __skb_queue_purge(&tsk->mc_method.deferredq);
633 sk_stop_timer(sk, &sk->sk_timer);
637 /* Reject any messages that accumulated in backlog queue */
639 tipc_dest_list_purge(&tsk->cong_links);
640 tsk->cong_link_cnt = 0;
641 call_rcu(&tsk->rcu, tipc_sk_callback);
648 * tipc_bind - associate or disassocate TIPC name(s) with a socket
649 * @sock: socket structure
650 * @uaddr: socket address describing name(s) and desired operation
651 * @uaddr_len: size of socket address data structure
653 * Name and name sequence binding is indicated using a positive scope value;
654 * a negative scope value unbinds the specified name. Specifying no name
655 * (i.e. a socket address length of 0) unbinds all names from the socket.
657 * Returns 0 on success, errno otherwise
659 * NOTE: This routine doesn't need to take the socket lock since it doesn't
660 * access any non-constant socket information.
662 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
665 struct sock *sk = sock->sk;
666 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
667 struct tipc_sock *tsk = tipc_sk(sk);
671 if (unlikely(!uaddr_len)) {
672 res = tipc_sk_withdraw(tsk, 0, NULL);
679 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
683 if (addr->family != AF_TIPC) {
688 if (addr->addrtype == TIPC_ADDR_NAME)
689 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
690 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
695 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
696 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
697 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
702 res = (addr->scope >= 0) ?
703 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
704 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
711 * tipc_getname - get port ID of socket or peer socket
712 * @sock: socket structure
713 * @uaddr: area for returned socket address
714 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
716 * Returns 0 on success, errno otherwise
718 * NOTE: This routine doesn't need to take the socket lock since it only
719 * accesses socket information that is unchanging (or which changes in
720 * a completely predictable manner).
722 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
725 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
726 struct sock *sk = sock->sk;
727 struct tipc_sock *tsk = tipc_sk(sk);
729 memset(addr, 0, sizeof(*addr));
731 if ((!tipc_sk_connected(sk)) &&
732 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
734 addr->addr.id.ref = tsk_peer_port(tsk);
735 addr->addr.id.node = tsk_peer_node(tsk);
737 addr->addr.id.ref = tsk->portid;
738 addr->addr.id.node = tipc_own_addr(sock_net(sk));
741 addr->addrtype = TIPC_ADDR_ID;
742 addr->family = AF_TIPC;
744 addr->addr.name.domain = 0;
746 return sizeof(*addr);
750 * tipc_poll - read and possibly block on pollmask
751 * @file: file structure associated with the socket
752 * @sock: socket for which to calculate the poll bits
755 * Returns pollmask value
758 * It appears that the usual socket locking mechanisms are not useful here
759 * since the pollmask info is potentially out-of-date the moment this routine
760 * exits. TCP and other protocols seem to rely on higher level poll routines
761 * to handle any preventable race conditions, so TIPC will do the same ...
763 * IMPORTANT: The fact that a read or write operation is indicated does NOT
764 * imply that the operation will succeed, merely that it should be performed
765 * and will not block.
767 static __poll_t tipc_poll(struct file *file, struct socket *sock,
770 struct sock *sk = sock->sk;
771 struct tipc_sock *tsk = tipc_sk(sk);
772 __poll_t revents = 0;
774 sock_poll_wait(file, sock, wait);
775 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
777 if (sk->sk_shutdown & RCV_SHUTDOWN)
778 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
779 if (sk->sk_shutdown == SHUTDOWN_MASK)
782 switch (sk->sk_state) {
783 case TIPC_ESTABLISHED:
784 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
788 case TIPC_CONNECTING:
789 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
790 revents |= EPOLLIN | EPOLLRDNORM;
793 if (tsk->group_is_open && !tsk->cong_link_cnt)
795 if (!tipc_sk_type_connectionless(sk))
797 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
799 revents |= EPOLLIN | EPOLLRDNORM;
801 case TIPC_DISCONNECTING:
802 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
809 * tipc_sendmcast - send multicast message
810 * @sock: socket structure
811 * @seq: destination address
812 * @msg: message to send
813 * @dlen: length of data to send
814 * @timeout: timeout to wait for wakeup
816 * Called from function tipc_sendmsg(), which has done all sanity checks
817 * Returns the number of bytes sent on success, or errno
819 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
820 struct msghdr *msg, size_t dlen, long timeout)
822 struct sock *sk = sock->sk;
823 struct tipc_sock *tsk = tipc_sk(sk);
824 struct tipc_msg *hdr = &tsk->phdr;
825 struct net *net = sock_net(sk);
826 int mtu = tipc_bcast_get_mtu(net);
827 struct tipc_mc_method *method = &tsk->mc_method;
828 struct sk_buff_head pkts;
829 struct tipc_nlist dsts;
835 /* Block or return if any destination link is congested */
836 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
840 /* Lookup destination nodes */
841 tipc_nlist_init(&dsts, tipc_own_addr(net));
842 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
844 if (!dsts.local && !dsts.remote)
845 return -EHOSTUNREACH;
847 /* Build message header */
848 msg_set_type(hdr, TIPC_MCAST_MSG);
849 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
850 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
851 msg_set_destport(hdr, 0);
852 msg_set_destnode(hdr, 0);
853 msg_set_nametype(hdr, seq->type);
854 msg_set_namelower(hdr, seq->lower);
855 msg_set_nameupper(hdr, seq->upper);
857 /* Build message as chain of buffers */
858 __skb_queue_head_init(&pkts);
859 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
861 /* Send message if build was successful */
862 if (unlikely(rc == dlen)) {
863 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
864 TIPC_DUMP_SK_SNDQ, " ");
865 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
866 &tsk->cong_link_cnt);
869 tipc_nlist_purge(&dsts);
871 return rc ? rc : dlen;
875 * tipc_send_group_msg - send a message to a member in the group
876 * @net: network namespace
877 * @m: message to send
879 * @dnode: destination node
880 * @dport: destination port
881 * @dlen: total length of message data
883 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
884 struct msghdr *m, struct tipc_member *mb,
885 u32 dnode, u32 dport, int dlen)
887 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
888 struct tipc_mc_method *method = &tsk->mc_method;
889 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
890 struct tipc_msg *hdr = &tsk->phdr;
891 struct sk_buff_head pkts;
894 /* Complete message header */
895 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
896 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
897 msg_set_destport(hdr, dport);
898 msg_set_destnode(hdr, dnode);
899 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
901 /* Build message as chain of buffers */
902 __skb_queue_head_init(&pkts);
903 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
904 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
905 if (unlikely(rc != dlen))
909 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
910 if (unlikely(rc == -ELINKCONG)) {
911 tipc_dest_push(&tsk->cong_links, dnode, 0);
912 tsk->cong_link_cnt++;
915 /* Update send window */
916 tipc_group_update_member(mb, blks);
918 /* A broadcast sent within next EXPIRE period must follow same path */
919 method->rcast = true;
920 method->mandatory = true;
925 * tipc_send_group_unicast - send message to a member in the group
926 * @sock: socket structure
927 * @m: message to send
928 * @dlen: total length of message data
929 * @timeout: timeout to wait for wakeup
931 * Called from function tipc_sendmsg(), which has done all sanity checks
932 * Returns the number of bytes sent on success, or errno
934 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
935 int dlen, long timeout)
937 struct sock *sk = sock->sk;
938 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
939 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
940 struct tipc_sock *tsk = tipc_sk(sk);
941 struct net *net = sock_net(sk);
942 struct tipc_member *mb = NULL;
946 node = dest->addr.id.node;
947 port = dest->addr.id.ref;
949 return -EHOSTUNREACH;
951 /* Block or return if destination link or member is congested */
952 rc = tipc_wait_for_cond(sock, &timeout,
953 !tipc_dest_find(&tsk->cong_links, node, 0) &&
955 !tipc_group_cong(tsk->group, node, port, blks,
961 return -EHOSTUNREACH;
963 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
965 return rc ? rc : dlen;
969 * tipc_send_group_anycast - send message to any member with given identity
970 * @sock: socket structure
971 * @m: message to send
972 * @dlen: total length of message data
973 * @timeout: timeout to wait for wakeup
975 * Called from function tipc_sendmsg(), which has done all sanity checks
976 * Returns the number of bytes sent on success, or errno
978 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
979 int dlen, long timeout)
981 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
982 struct sock *sk = sock->sk;
983 struct tipc_sock *tsk = tipc_sk(sk);
984 struct list_head *cong_links = &tsk->cong_links;
985 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
986 struct tipc_msg *hdr = &tsk->phdr;
987 struct tipc_member *first = NULL;
988 struct tipc_member *mbr = NULL;
989 struct net *net = sock_net(sk);
990 u32 node, port, exclude;
991 struct list_head dsts;
992 u32 type, inst, scope;
997 INIT_LIST_HEAD(&dsts);
999 type = msg_nametype(hdr);
1000 inst = dest->addr.name.name.instance;
1001 scope = msg_lookup_scope(hdr);
1003 while (++lookups < 4) {
1004 exclude = tipc_group_exclude(tsk->group);
1008 /* Look for a non-congested destination member, if any */
1010 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1011 &dstcnt, exclude, false))
1012 return -EHOSTUNREACH;
1013 tipc_dest_pop(&dsts, &node, &port);
1014 cong = tipc_group_cong(tsk->group, node, port, blks,
1024 /* Start over if destination was not in member list */
1028 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1031 /* Block or return if destination link or member is congested */
1032 rc = tipc_wait_for_cond(sock, &timeout,
1033 !tipc_dest_find(cong_links, node, 0) &&
1035 !tipc_group_cong(tsk->group, node, port,
1040 /* Send, unless destination disappeared while waiting */
1045 if (unlikely(lookups >= 4))
1046 return -EHOSTUNREACH;
1048 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1050 return rc ? rc : dlen;
1054 * tipc_send_group_bcast - send message to all members in communication group
1055 * @sock: socket structure
1056 * @m: message to send
1057 * @dlen: total length of message data
1058 * @timeout: timeout to wait for wakeup
1060 * Called from function tipc_sendmsg(), which has done all sanity checks
1061 * Returns the number of bytes sent on success, or errno
1063 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1064 int dlen, long timeout)
1066 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1067 struct sock *sk = sock->sk;
1068 struct net *net = sock_net(sk);
1069 struct tipc_sock *tsk = tipc_sk(sk);
1070 struct tipc_nlist *dsts;
1071 struct tipc_mc_method *method = &tsk->mc_method;
1072 bool ack = method->mandatory && method->rcast;
1073 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1074 struct tipc_msg *hdr = &tsk->phdr;
1075 int mtu = tipc_bcast_get_mtu(net);
1076 struct sk_buff_head pkts;
1077 int rc = -EHOSTUNREACH;
1079 /* Block or return if any destination link or member is congested */
1080 rc = tipc_wait_for_cond(sock, &timeout,
1081 !tsk->cong_link_cnt && tsk->group &&
1082 !tipc_group_bc_cong(tsk->group, blks));
1086 dsts = tipc_group_dests(tsk->group);
1087 if (!dsts->local && !dsts->remote)
1088 return -EHOSTUNREACH;
1090 /* Complete message header */
1092 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1093 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1095 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1096 msg_set_nameinst(hdr, 0);
1098 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1099 msg_set_destport(hdr, 0);
1100 msg_set_destnode(hdr, 0);
1101 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1103 /* Avoid getting stuck with repeated forced replicasts */
1104 msg_set_grp_bc_ack_req(hdr, ack);
1106 /* Build message as chain of buffers */
1107 __skb_queue_head_init(&pkts);
1108 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1109 if (unlikely(rc != dlen))
1113 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1117 /* Update broadcast sequence number and send windows */
1118 tipc_group_update_bc_members(tsk->group, blks, ack);
1120 /* Broadcast link is now free to choose method for next broadcast */
1121 method->mandatory = false;
1122 method->expires = jiffies;
1128 * tipc_send_group_mcast - send message to all members with given identity
1129 * @sock: socket structure
1130 * @m: message to send
1131 * @dlen: total length of message data
1132 * @timeout: timeout to wait for wakeup
1134 * Called from function tipc_sendmsg(), which has done all sanity checks
1135 * Returns the number of bytes sent on success, or errno
1137 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1138 int dlen, long timeout)
1140 struct sock *sk = sock->sk;
1141 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1142 struct tipc_sock *tsk = tipc_sk(sk);
1143 struct tipc_group *grp = tsk->group;
1144 struct tipc_msg *hdr = &tsk->phdr;
1145 struct net *net = sock_net(sk);
1146 u32 type, inst, scope, exclude;
1147 struct list_head dsts;
1150 INIT_LIST_HEAD(&dsts);
1152 type = msg_nametype(hdr);
1153 inst = dest->addr.name.name.instance;
1154 scope = msg_lookup_scope(hdr);
1155 exclude = tipc_group_exclude(grp);
1157 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1158 &dstcnt, exclude, true))
1159 return -EHOSTUNREACH;
1162 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1163 return tipc_send_group_unicast(sock, m, dlen, timeout);
1166 tipc_dest_list_purge(&dsts);
1167 return tipc_send_group_bcast(sock, m, dlen, timeout);
1171 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1172 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1173 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1175 * Multi-threaded: parallel calls with reference to same queues may occur
1177 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1178 struct sk_buff_head *inputq)
1180 u32 self = tipc_own_addr(net);
1181 u32 type, lower, upper, scope;
1182 struct sk_buff *skb, *_skb;
1184 struct sk_buff_head tmpq;
1185 struct list_head dports;
1186 struct tipc_msg *hdr;
1187 int user, mtyp, hlen;
1190 __skb_queue_head_init(&tmpq);
1191 INIT_LIST_HEAD(&dports);
1193 skb = tipc_skb_peek(arrvq, &inputq->lock);
1194 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1196 user = msg_user(hdr);
1197 mtyp = msg_type(hdr);
1198 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1199 onode = msg_orignode(hdr);
1200 type = msg_nametype(hdr);
1202 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1203 spin_lock_bh(&inputq->lock);
1204 if (skb_peek(arrvq) == skb) {
1205 __skb_dequeue(arrvq);
1206 __skb_queue_tail(inputq, skb);
1209 spin_unlock_bh(&inputq->lock);
1213 /* Group messages require exact scope match */
1214 if (msg_in_group(hdr)) {
1217 scope = msg_lookup_scope(hdr);
1220 /* TIPC_NODE_SCOPE means "any scope" in this context */
1222 scope = TIPC_NODE_SCOPE;
1224 scope = TIPC_CLUSTER_SCOPE;
1226 lower = msg_namelower(hdr);
1227 upper = msg_nameupper(hdr);
1230 /* Create destination port list: */
1231 tipc_nametbl_mc_lookup(net, type, lower, upper,
1232 scope, exact, &dports);
1234 /* Clone message per destination */
1235 while (tipc_dest_pop(&dports, NULL, &portid)) {
1236 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1238 msg_set_destport(buf_msg(_skb), portid);
1239 __skb_queue_tail(&tmpq, _skb);
1242 pr_warn("Failed to clone mcast rcv buffer\n");
1244 /* Append to inputq if not already done by other thread */
1245 spin_lock_bh(&inputq->lock);
1246 if (skb_peek(arrvq) == skb) {
1247 skb_queue_splice_tail_init(&tmpq, inputq);
1248 kfree_skb(__skb_dequeue(arrvq));
1250 spin_unlock_bh(&inputq->lock);
1251 __skb_queue_purge(&tmpq);
1254 tipc_sk_rcv(net, inputq);
1257 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1258 * when socket is in Nagle mode
1260 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1262 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1263 struct sk_buff *skb = skb_peek_tail(txq);
1264 struct net *net = sock_net(&tsk->sk);
1265 u32 dnode = tsk_peer_node(tsk);
1269 tsk->pkt_cnt += skb_queue_len(txq);
1270 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1272 if (tsk->nagle_start < NAGLE_START_MAX)
1273 tsk->nagle_start *= 2;
1274 tsk->expect_ack = false;
1275 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1276 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1279 tsk->nagle_start = NAGLE_START_INIT;
1281 msg_set_ack_required(buf_msg(skb));
1282 tsk->expect_ack = true;
1284 tsk->expect_ack = false;
1291 if (!skb || tsk->cong_link_cnt)
1294 /* Do not send SYN again after congestion */
1295 if (msg_is_syn(buf_msg(skb)))
1299 tsk->pkt_cnt += skb_queue_len(txq);
1300 tsk->snt_unacked += tsk->snd_backlog;
1301 tsk->snd_backlog = 0;
1302 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1303 if (rc == -ELINKCONG)
1304 tsk->cong_link_cnt = 1;
1308 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1309 * @tsk: receiving socket
1310 * @skb: pointer to message buffer.
1312 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1313 struct sk_buff_head *inputq,
1314 struct sk_buff_head *xmitq)
1316 struct tipc_msg *hdr = buf_msg(skb);
1317 u32 onode = tsk_own_node(tsk);
1318 struct sock *sk = &tsk->sk;
1319 int mtyp = msg_type(hdr);
1322 /* Ignore if connection cannot be validated: */
1323 if (!tsk_peer_msg(tsk, hdr)) {
1324 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1328 if (unlikely(msg_errcode(hdr))) {
1329 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1330 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1331 tsk_peer_port(tsk));
1332 sk->sk_state_change(sk);
1334 /* State change is ignored if socket already awake,
1335 * - convert msg to abort msg and add to inqueue
1337 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1338 msg_set_type(hdr, TIPC_CONN_MSG);
1339 msg_set_size(hdr, BASIC_H_SIZE);
1340 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1341 __skb_queue_tail(inputq, skb);
1345 tsk->probe_unacked = false;
1347 if (mtyp == CONN_PROBE) {
1348 msg_set_type(hdr, CONN_PROBE_REPLY);
1349 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1350 __skb_queue_tail(xmitq, skb);
1352 } else if (mtyp == CONN_ACK) {
1353 was_cong = tsk_conn_cong(tsk);
1354 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1355 tsk->snt_unacked -= msg_conn_ack(hdr);
1356 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1357 tsk->snd_win = msg_adv_win(hdr);
1358 if (was_cong && !tsk_conn_cong(tsk))
1359 sk->sk_write_space(sk);
1360 } else if (mtyp != CONN_PROBE_REPLY) {
1361 pr_warn("Received unknown CONN_PROTO msg\n");
1368 * tipc_sendmsg - send message in connectionless manner
1369 * @sock: socket structure
1370 * @m: message to send
1371 * @dsz: amount of user data to be sent
1373 * Message must have an destination specified explicitly.
1374 * Used for SOCK_RDM and SOCK_DGRAM messages,
1375 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1376 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1378 * Returns the number of bytes sent on success, or errno otherwise
1380 static int tipc_sendmsg(struct socket *sock,
1381 struct msghdr *m, size_t dsz)
1383 struct sock *sk = sock->sk;
1387 ret = __tipc_sendmsg(sock, m, dsz);
1393 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1395 struct sock *sk = sock->sk;
1396 struct net *net = sock_net(sk);
1397 struct tipc_sock *tsk = tipc_sk(sk);
1398 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1399 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1400 struct list_head *clinks = &tsk->cong_links;
1401 bool syn = !tipc_sk_type_connectionless(sk);
1402 struct tipc_group *grp = tsk->group;
1403 struct tipc_msg *hdr = &tsk->phdr;
1404 struct tipc_name_seq *seq;
1405 struct sk_buff_head pkts;
1406 u32 dport = 0, dnode = 0;
1407 u32 type = 0, inst = 0;
1410 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1414 if (unlikely(m->msg_namelen < sizeof(*dest)))
1416 if (unlikely(dest->family != AF_TIPC))
1422 return tipc_send_group_bcast(sock, m, dlen, timeout);
1423 if (dest->addrtype == TIPC_ADDR_NAME)
1424 return tipc_send_group_anycast(sock, m, dlen, timeout);
1425 if (dest->addrtype == TIPC_ADDR_ID)
1426 return tipc_send_group_unicast(sock, m, dlen, timeout);
1427 if (dest->addrtype == TIPC_ADDR_MCAST)
1428 return tipc_send_group_mcast(sock, m, dlen, timeout);
1432 if (unlikely(!dest)) {
1434 if (!syn && dest->family != AF_TIPC)
1435 return -EDESTADDRREQ;
1438 if (unlikely(syn)) {
1439 if (sk->sk_state == TIPC_LISTEN)
1441 if (sk->sk_state != TIPC_OPEN)
1445 if (dest->addrtype == TIPC_ADDR_NAME) {
1446 tsk->conn_type = dest->addr.name.name.type;
1447 tsk->conn_instance = dest->addr.name.name.instance;
1449 msg_set_syn(hdr, 1);
1452 seq = &dest->addr.nameseq;
1453 if (dest->addrtype == TIPC_ADDR_MCAST)
1454 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1456 if (dest->addrtype == TIPC_ADDR_NAME) {
1457 type = dest->addr.name.name.type;
1458 inst = dest->addr.name.name.instance;
1459 dnode = dest->addr.name.domain;
1460 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1461 if (unlikely(!dport && !dnode))
1462 return -EHOSTUNREACH;
1463 } else if (dest->addrtype == TIPC_ADDR_ID) {
1464 dnode = dest->addr.id.node;
1469 /* Block or return if destination link is congested */
1470 rc = tipc_wait_for_cond(sock, &timeout,
1471 !tipc_dest_find(clinks, dnode, 0));
1475 if (dest->addrtype == TIPC_ADDR_NAME) {
1476 msg_set_type(hdr, TIPC_NAMED_MSG);
1477 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1478 msg_set_nametype(hdr, type);
1479 msg_set_nameinst(hdr, inst);
1480 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1481 msg_set_destnode(hdr, dnode);
1482 msg_set_destport(hdr, dport);
1483 } else { /* TIPC_ADDR_ID */
1484 msg_set_type(hdr, TIPC_DIRECT_MSG);
1485 msg_set_lookup_scope(hdr, 0);
1486 msg_set_destnode(hdr, dnode);
1487 msg_set_destport(hdr, dest->addr.id.ref);
1488 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1491 __skb_queue_head_init(&pkts);
1492 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, true);
1493 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1494 if (unlikely(rc != dlen))
1496 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1497 __skb_queue_purge(&pkts);
1501 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1502 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1503 if (unlikely(rc == -ELINKCONG)) {
1504 tipc_dest_push(clinks, dnode, 0);
1505 tsk->cong_link_cnt++;
1509 if (unlikely(syn && !rc))
1510 tipc_set_sk_state(sk, TIPC_CONNECTING);
1512 return rc ? rc : dlen;
1516 * tipc_sendstream - send stream-oriented data
1517 * @sock: socket structure
1519 * @dsz: total length of data to be transmitted
1521 * Used for SOCK_STREAM data.
1523 * Returns the number of bytes sent on success (or partial success),
1524 * or errno if no data sent
1526 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1528 struct sock *sk = sock->sk;
1532 ret = __tipc_sendstream(sock, m, dsz);
1538 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1540 struct sock *sk = sock->sk;
1541 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1542 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1543 struct sk_buff_head *txq = &sk->sk_write_queue;
1544 struct tipc_sock *tsk = tipc_sk(sk);
1545 struct tipc_msg *hdr = &tsk->phdr;
1546 struct net *net = sock_net(sk);
1547 struct sk_buff *skb;
1548 u32 dnode = tsk_peer_node(tsk);
1549 int maxnagle = tsk->maxnagle;
1550 int maxpkt = tsk->max_pkt;
1554 if (unlikely(dlen > INT_MAX))
1557 /* Handle implicit connection setup */
1558 if (unlikely(dest)) {
1559 rc = __tipc_sendmsg(sock, m, dlen);
1560 if (dlen && dlen == rc) {
1561 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1562 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1568 rc = tipc_wait_for_cond(sock, &timeout,
1569 (!tsk->cong_link_cnt &&
1570 !tsk_conn_cong(tsk) &&
1571 tipc_sk_connected(sk)));
1574 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1575 blocks = tsk->snd_backlog;
1576 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1578 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1579 if (unlikely(rc < 0))
1583 if (blocks <= 64 && tsk->expect_ack) {
1584 tsk->snd_backlog = blocks;
1587 } else if (blocks > 64) {
1588 tsk->pkt_cnt += skb_queue_len(txq);
1590 skb = skb_peek_tail(txq);
1592 msg_set_ack_required(buf_msg(skb));
1593 tsk->expect_ack = true;
1595 tsk->expect_ack = false;
1601 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1602 if (unlikely(rc != send))
1604 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1606 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1607 TIPC_DUMP_SK_SNDQ, " ");
1608 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1609 if (unlikely(rc == -ELINKCONG)) {
1610 tsk->cong_link_cnt = 1;
1614 tsk->snt_unacked += blocks;
1615 tsk->snd_backlog = 0;
1618 } while (sent < dlen && !rc);
1620 return sent ? sent : rc;
1624 * tipc_send_packet - send a connection-oriented message
1625 * @sock: socket structure
1626 * @m: message to send
1627 * @dsz: length of data to be transmitted
1629 * Used for SOCK_SEQPACKET messages.
1631 * Returns the number of bytes sent on success, or errno otherwise
1633 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1635 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1638 return tipc_sendstream(sock, m, dsz);
1641 /* tipc_sk_finish_conn - complete the setup of a connection
1643 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1646 struct sock *sk = &tsk->sk;
1647 struct net *net = sock_net(sk);
1648 struct tipc_msg *msg = &tsk->phdr;
1650 msg_set_syn(msg, 0);
1651 msg_set_destnode(msg, peer_node);
1652 msg_set_destport(msg, peer_port);
1653 msg_set_type(msg, TIPC_CONN_MSG);
1654 msg_set_lookup_scope(msg, 0);
1655 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1657 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1658 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1659 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1660 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1661 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1663 __skb_queue_purge(&sk->sk_write_queue);
1664 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1667 /* Fall back to message based flow control */
1668 tsk->rcv_win = FLOWCTL_MSG_WIN;
1669 tsk->snd_win = FLOWCTL_MSG_WIN;
1673 * tipc_sk_set_orig_addr - capture sender's address for received message
1674 * @m: descriptor for message info
1675 * @skb: received message
1677 * Note: Address is not captured if not requested by receiver.
1679 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1681 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1682 struct tipc_msg *hdr = buf_msg(skb);
1687 srcaddr->sock.family = AF_TIPC;
1688 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1689 srcaddr->sock.scope = 0;
1690 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1691 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1692 srcaddr->sock.addr.name.domain = 0;
1693 m->msg_namelen = sizeof(struct sockaddr_tipc);
1695 if (!msg_in_group(hdr))
1698 /* Group message users may also want to know sending member's id */
1699 srcaddr->member.family = AF_TIPC;
1700 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1701 srcaddr->member.scope = 0;
1702 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1703 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1704 srcaddr->member.addr.name.domain = 0;
1705 m->msg_namelen = sizeof(*srcaddr);
1709 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1710 * @m: descriptor for message info
1711 * @skb: received message buffer
1712 * @tsk: TIPC port associated with message
1714 * Note: Ancillary data is not captured if not requested by receiver.
1716 * Returns 0 if successful, otherwise errno
1718 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1719 struct tipc_sock *tsk)
1721 struct tipc_msg *msg;
1728 if (likely(m->msg_controllen == 0))
1732 /* Optionally capture errored message object(s) */
1733 err = msg ? msg_errcode(msg) : 0;
1734 if (unlikely(err)) {
1736 anc_data[1] = msg_data_sz(msg);
1737 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1741 if (skb_linearize(skb))
1744 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1751 /* Optionally capture message destination object */
1752 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1753 switch (dest_type) {
1754 case TIPC_NAMED_MSG:
1756 anc_data[0] = msg_nametype(msg);
1757 anc_data[1] = msg_namelower(msg);
1758 anc_data[2] = msg_namelower(msg);
1760 case TIPC_MCAST_MSG:
1762 anc_data[0] = msg_nametype(msg);
1763 anc_data[1] = msg_namelower(msg);
1764 anc_data[2] = msg_nameupper(msg);
1767 has_name = (tsk->conn_type != 0);
1768 anc_data[0] = tsk->conn_type;
1769 anc_data[1] = tsk->conn_instance;
1770 anc_data[2] = tsk->conn_instance;
1776 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1784 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1786 struct sock *sk = &tsk->sk;
1787 struct sk_buff *skb = NULL;
1788 struct tipc_msg *msg;
1789 u32 peer_port = tsk_peer_port(tsk);
1790 u32 dnode = tsk_peer_node(tsk);
1792 if (!tipc_sk_connected(sk))
1794 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1795 dnode, tsk_own_node(tsk), peer_port,
1796 tsk->portid, TIPC_OK);
1800 msg_set_conn_ack(msg, tsk->rcv_unacked);
1801 tsk->rcv_unacked = 0;
1803 /* Adjust to and advertize the correct window limit */
1804 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1805 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1806 msg_set_adv_win(msg, tsk->rcv_win);
1811 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1813 struct sk_buff *skb;
1815 skb = tipc_sk_build_ack(tsk);
1819 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1820 msg_link_selector(buf_msg(skb)));
1823 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1825 struct sock *sk = sock->sk;
1826 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1827 long timeo = *timeop;
1828 int err = sock_error(sk);
1834 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1835 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1839 add_wait_queue(sk_sleep(sk), &wait);
1841 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1842 sched_annotate_sleep();
1844 remove_wait_queue(sk_sleep(sk), &wait);
1847 if (!skb_queue_empty(&sk->sk_receive_queue))
1852 err = sock_intr_errno(timeo);
1853 if (signal_pending(current))
1856 err = sock_error(sk);
1865 * tipc_recvmsg - receive packet-oriented message
1866 * @m: descriptor for message info
1867 * @buflen: length of user buffer area
1868 * @flags: receive flags
1870 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1871 * If the complete message doesn't fit in user area, truncate it.
1873 * Returns size of returned message data, errno otherwise
1875 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1876 size_t buflen, int flags)
1878 struct sock *sk = sock->sk;
1879 bool connected = !tipc_sk_type_connectionless(sk);
1880 struct tipc_sock *tsk = tipc_sk(sk);
1881 int rc, err, hlen, dlen, copy;
1882 struct sk_buff_head xmitq;
1883 struct tipc_msg *hdr;
1884 struct sk_buff *skb;
1888 /* Catch invalid receive requests */
1889 if (unlikely(!buflen))
1893 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1897 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1899 /* Step rcv queue to first msg with data or error; wait if necessary */
1901 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1904 skb = skb_peek(&sk->sk_receive_queue);
1906 dlen = msg_data_sz(hdr);
1907 hlen = msg_hdr_sz(hdr);
1908 err = msg_errcode(hdr);
1909 grp_evt = msg_is_grp_evt(hdr);
1910 if (likely(dlen || err))
1912 tsk_advance_rx_queue(sk);
1915 /* Collect msg meta data, including error code and rejected data */
1916 tipc_sk_set_orig_addr(m, skb);
1917 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1922 /* Capture data if non-error msg, otherwise just set return value */
1924 copy = min_t(int, dlen, buflen);
1925 if (unlikely(copy != dlen))
1926 m->msg_flags |= MSG_TRUNC;
1927 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1931 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1937 /* Mark message as group event if applicable */
1938 if (unlikely(grp_evt)) {
1939 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1940 m->msg_flags |= MSG_EOR;
1941 m->msg_flags |= MSG_OOB;
1945 /* Caption of data or error code/rejected data was successful */
1946 if (unlikely(flags & MSG_PEEK))
1949 /* Send group flow control advertisement when applicable */
1950 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1951 __skb_queue_head_init(&xmitq);
1952 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1953 msg_orignode(hdr), msg_origport(hdr),
1955 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1958 tsk_advance_rx_queue(sk);
1960 if (likely(!connected))
1963 /* Send connection flow control advertisement when applicable */
1964 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1965 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1966 tipc_sk_send_ack(tsk);
1969 return rc ? rc : copy;
1973 * tipc_recvstream - receive stream-oriented data
1974 * @m: descriptor for message info
1975 * @buflen: total size of user buffer area
1976 * @flags: receive flags
1978 * Used for SOCK_STREAM messages only. If not enough data is available
1979 * will optionally wait for more; never truncates data.
1981 * Returns size of returned message data, errno otherwise
1983 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1984 size_t buflen, int flags)
1986 struct sock *sk = sock->sk;
1987 struct tipc_sock *tsk = tipc_sk(sk);
1988 struct sk_buff *skb;
1989 struct tipc_msg *hdr;
1990 struct tipc_skb_cb *skb_cb;
1991 bool peek = flags & MSG_PEEK;
1992 int offset, required, copy, copied = 0;
1993 int hlen, dlen, err, rc;
1996 /* Catch invalid receive attempts */
1997 if (unlikely(!buflen))
2002 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2006 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2007 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2010 /* Look at first msg in receive queue; wait if necessary */
2011 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2014 skb = skb_peek(&sk->sk_receive_queue);
2015 skb_cb = TIPC_SKB_CB(skb);
2017 dlen = msg_data_sz(hdr);
2018 hlen = msg_hdr_sz(hdr);
2019 err = msg_errcode(hdr);
2021 /* Discard any empty non-errored (SYN-) message */
2022 if (unlikely(!dlen && !err)) {
2023 tsk_advance_rx_queue(sk);
2027 /* Collect msg meta data, incl. error code and rejected data */
2029 tipc_sk_set_orig_addr(m, skb);
2030 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2036 /* Copy data if msg ok, otherwise return error/partial data */
2038 offset = skb_cb->bytes_read;
2039 copy = min_t(int, dlen - offset, buflen - copied);
2040 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2045 if (unlikely(offset < dlen)) {
2047 skb_cb->bytes_read = offset;
2052 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2061 tsk_advance_rx_queue(sk);
2063 /* Send connection flow control advertisement when applicable */
2064 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2065 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2066 tipc_sk_send_ack(tsk);
2068 /* Exit if all requested data or FIN/error received */
2069 if (copied == buflen || err)
2072 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2075 return copied ? copied : rc;
2079 * tipc_write_space - wake up thread if port congestion is released
2082 static void tipc_write_space(struct sock *sk)
2084 struct socket_wq *wq;
2087 wq = rcu_dereference(sk->sk_wq);
2088 if (skwq_has_sleeper(wq))
2089 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2090 EPOLLWRNORM | EPOLLWRBAND);
2095 * tipc_data_ready - wake up threads to indicate messages have been received
2098 static void tipc_data_ready(struct sock *sk)
2100 struct socket_wq *wq;
2103 wq = rcu_dereference(sk->sk_wq);
2104 if (skwq_has_sleeper(wq))
2105 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2106 EPOLLRDNORM | EPOLLRDBAND);
2110 static void tipc_sock_destruct(struct sock *sk)
2112 __skb_queue_purge(&sk->sk_receive_queue);
2115 static void tipc_sk_proto_rcv(struct sock *sk,
2116 struct sk_buff_head *inputq,
2117 struct sk_buff_head *xmitq)
2119 struct sk_buff *skb = __skb_dequeue(inputq);
2120 struct tipc_sock *tsk = tipc_sk(sk);
2121 struct tipc_msg *hdr = buf_msg(skb);
2122 struct tipc_group *grp = tsk->group;
2123 bool wakeup = false;
2125 switch (msg_user(hdr)) {
2127 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2130 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2131 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2133 tsk->cong_link_cnt--;
2135 tipc_sk_push_backlog(tsk, false);
2137 case GROUP_PROTOCOL:
2138 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2141 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2142 hdr, inputq, xmitq);
2149 sk->sk_write_space(sk);
2155 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2157 * @skb: pointer to message buffer.
2158 * @xmitq: for Nagle ACK if any
2159 * Returns true if message should be added to receive queue, false otherwise
2161 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2162 struct sk_buff_head *xmitq)
2164 struct sock *sk = &tsk->sk;
2165 struct net *net = sock_net(sk);
2166 struct tipc_msg *hdr = buf_msg(skb);
2167 bool con_msg = msg_connected(hdr);
2168 u32 pport = tsk_peer_port(tsk);
2169 u32 pnode = tsk_peer_node(tsk);
2170 u32 oport = msg_origport(hdr);
2171 u32 onode = msg_orignode(hdr);
2172 int err = msg_errcode(hdr);
2173 unsigned long delay;
2175 if (unlikely(msg_mcast(hdr)))
2179 switch (sk->sk_state) {
2180 case TIPC_CONNECTING:
2182 if (likely(con_msg)) {
2185 tipc_sk_finish_conn(tsk, oport, onode);
2186 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2187 /* ACK+ message with data is added to receive queue */
2188 if (msg_data_sz(hdr))
2190 /* Empty ACK-, - wake up sleeping connect() and drop */
2191 sk->sk_state_change(sk);
2192 msg_set_dest_droppable(hdr, 1);
2195 /* Ignore connectionless message if not from listening socket */
2196 if (oport != pport || onode != pnode)
2200 if (err != TIPC_ERR_OVERLOAD)
2203 /* Prepare for new setup attempt if we have a SYN clone */
2204 if (skb_queue_empty(&sk->sk_write_queue))
2206 get_random_bytes(&delay, 2);
2207 delay %= (tsk->conn_timeout / 4);
2208 delay = msecs_to_jiffies(delay + 100);
2209 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2212 case TIPC_DISCONNECTING:
2215 /* Accept only SYN message */
2216 if (!msg_is_syn(hdr) &&
2217 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2219 if (!con_msg && !err)
2222 case TIPC_ESTABLISHED:
2223 if (!skb_queue_empty(&sk->sk_write_queue))
2224 tipc_sk_push_backlog(tsk, false);
2225 /* Accept only connection-based messages sent by peer */
2226 if (likely(con_msg && !err && pport == oport &&
2228 if (msg_ack_required(hdr)) {
2229 struct sk_buff *skb;
2231 skb = tipc_sk_build_ack(tsk);
2233 msg_set_nagle_ack(buf_msg(skb));
2234 __skb_queue_tail(xmitq, skb);
2239 if (!tsk_peer_msg(tsk, hdr))
2243 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2244 tipc_node_remove_conn(net, pnode, tsk->portid);
2245 sk->sk_state_change(sk);
2248 pr_err("Unknown sk_state %u\n", sk->sk_state);
2250 /* Abort connection setup attempt */
2251 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2252 sk->sk_err = ECONNREFUSED;
2253 sk->sk_state_change(sk);
2258 * rcvbuf_limit - get proper overload limit of socket receive queue
2262 * For connection oriented messages, irrespective of importance,
2263 * default queue limit is 2 MB.
2265 * For connectionless messages, queue limits are based on message
2266 * importance as follows:
2268 * TIPC_LOW_IMPORTANCE (2 MB)
2269 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2270 * TIPC_HIGH_IMPORTANCE (8 MB)
2271 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2273 * Returns overload limit according to corresponding message importance
2275 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2277 struct tipc_sock *tsk = tipc_sk(sk);
2278 struct tipc_msg *hdr = buf_msg(skb);
2280 if (unlikely(msg_in_group(hdr)))
2281 return READ_ONCE(sk->sk_rcvbuf);
2283 if (unlikely(!msg_connected(hdr)))
2284 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2286 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2287 return READ_ONCE(sk->sk_rcvbuf);
2289 return FLOWCTL_MSG_LIM;
2293 * tipc_sk_filter_rcv - validate incoming message
2295 * @skb: pointer to message.
2297 * Enqueues message on receive queue if acceptable; optionally handles
2298 * disconnect indication for a connected socket.
2300 * Called with socket lock already taken
2303 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2304 struct sk_buff_head *xmitq)
2306 bool sk_conn = !tipc_sk_type_connectionless(sk);
2307 struct tipc_sock *tsk = tipc_sk(sk);
2308 struct tipc_group *grp = tsk->group;
2309 struct tipc_msg *hdr = buf_msg(skb);
2310 struct net *net = sock_net(sk);
2311 struct sk_buff_head inputq;
2312 int mtyp = msg_type(hdr);
2313 int limit, err = TIPC_OK;
2315 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2316 TIPC_SKB_CB(skb)->bytes_read = 0;
2317 __skb_queue_head_init(&inputq);
2318 __skb_queue_tail(&inputq, skb);
2320 if (unlikely(!msg_isdata(hdr)))
2321 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2324 tipc_group_filter_msg(grp, &inputq, xmitq);
2326 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2327 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2329 /* Validate and add to receive buffer if there is space */
2330 while ((skb = __skb_dequeue(&inputq))) {
2332 limit = rcvbuf_limit(sk, skb);
2333 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2334 (!sk_conn && msg_connected(hdr)) ||
2335 (!grp && msg_in_group(hdr)))
2336 err = TIPC_ERR_NO_PORT;
2337 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2338 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2340 atomic_inc(&sk->sk_drops);
2341 err = TIPC_ERR_OVERLOAD;
2344 if (unlikely(err)) {
2345 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2346 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2348 __skb_queue_tail(xmitq, skb);
2353 __skb_queue_tail(&sk->sk_receive_queue, skb);
2354 skb_set_owner_r(skb, sk);
2355 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2356 "rcvq >90% allocated!");
2357 sk->sk_data_ready(sk);
2362 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2366 * Caller must hold socket lock
2368 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2370 unsigned int before = sk_rmem_alloc_get(sk);
2371 struct sk_buff_head xmitq;
2374 __skb_queue_head_init(&xmitq);
2376 tipc_sk_filter_rcv(sk, skb, &xmitq);
2377 added = sk_rmem_alloc_get(sk) - before;
2378 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2380 /* Send pending response/rejected messages, if any */
2381 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2386 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2387 * inputq and try adding them to socket or backlog queue
2388 * @inputq: list of incoming buffers with potentially different destinations
2389 * @sk: socket where the buffers should be enqueued
2390 * @dport: port number for the socket
2392 * Caller must hold socket lock
2394 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2395 u32 dport, struct sk_buff_head *xmitq)
2397 unsigned long time_limit = jiffies + 2;
2398 struct sk_buff *skb;
2403 while (skb_queue_len(inputq)) {
2404 if (unlikely(time_after_eq(jiffies, time_limit)))
2407 skb = tipc_skb_dequeue(inputq, dport);
2411 /* Add message directly to receive queue if possible */
2412 if (!sock_owned_by_user(sk)) {
2413 tipc_sk_filter_rcv(sk, skb, xmitq);
2417 /* Try backlog, compensating for double-counted bytes */
2418 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2419 if (!sk->sk_backlog.len)
2420 atomic_set(dcnt, 0);
2421 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2422 if (likely(!sk_add_backlog(sk, skb, lim))) {
2423 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2424 "bklg & rcvq >90% allocated!");
2428 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2429 /* Overload => reject message back to sender */
2430 onode = tipc_own_addr(sock_net(sk));
2431 atomic_inc(&sk->sk_drops);
2432 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2433 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2435 __skb_queue_tail(xmitq, skb);
2442 * tipc_sk_rcv - handle a chain of incoming buffers
2443 * @inputq: buffer list containing the buffers
2444 * Consumes all buffers in list until inputq is empty
2445 * Note: may be called in multiple threads referring to the same queue
2447 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2449 struct sk_buff_head xmitq;
2450 u32 dnode, dport = 0;
2452 struct tipc_sock *tsk;
2454 struct sk_buff *skb;
2456 __skb_queue_head_init(&xmitq);
2457 while (skb_queue_len(inputq)) {
2458 dport = tipc_skb_peek_port(inputq, dport);
2459 tsk = tipc_sk_lookup(net, dport);
2463 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2464 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2465 spin_unlock_bh(&sk->sk_lock.slock);
2467 /* Send pending response/rejected messages, if any */
2468 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2472 /* No destination socket => dequeue skb if still there */
2473 skb = tipc_skb_dequeue(inputq, dport);
2477 /* Try secondary lookup if unresolved named message */
2478 err = TIPC_ERR_NO_PORT;
2479 if (tipc_msg_lookup_dest(net, skb, &err))
2482 /* Prepare for message rejection */
2483 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2486 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2488 dnode = msg_destnode(buf_msg(skb));
2489 tipc_node_xmit_skb(net, skb, dnode, dport);
2493 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2495 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2496 struct sock *sk = sock->sk;
2500 int err = sock_error(sk);
2505 if (signal_pending(current))
2506 return sock_intr_errno(*timeo_p);
2507 if (sk->sk_state == TIPC_DISCONNECTING)
2510 add_wait_queue(sk_sleep(sk), &wait);
2511 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2513 remove_wait_queue(sk_sleep(sk), &wait);
2518 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2520 if (addr->family != AF_TIPC)
2522 if (addr->addrtype == TIPC_SERVICE_RANGE)
2523 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2524 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2525 addr->addrtype == TIPC_SOCKET_ADDR);
2529 * tipc_connect - establish a connection to another TIPC port
2530 * @sock: socket structure
2531 * @dest: socket address for destination port
2532 * @destlen: size of socket address data structure
2533 * @flags: file-related flags associated with socket
2535 * Returns 0 on success, errno otherwise
2537 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2538 int destlen, int flags)
2540 struct sock *sk = sock->sk;
2541 struct tipc_sock *tsk = tipc_sk(sk);
2542 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2543 struct msghdr m = {NULL,};
2544 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2548 if (destlen != sizeof(struct sockaddr_tipc))
2558 if (dst->family == AF_UNSPEC) {
2559 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2560 if (!tipc_sk_type_connectionless(sk))
2564 if (!tipc_sockaddr_is_sane(dst)) {
2568 /* DGRAM/RDM connect(), just save the destaddr */
2569 if (tipc_sk_type_connectionless(sk)) {
2570 memcpy(&tsk->peer, dest, destlen);
2572 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2577 previous = sk->sk_state;
2579 switch (sk->sk_state) {
2581 /* Send a 'SYN-' to destination */
2583 m.msg_namelen = destlen;
2585 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2586 * indicate send_msg() is never blocked.
2589 m.msg_flags = MSG_DONTWAIT;
2591 res = __tipc_sendmsg(sock, &m, 0);
2592 if ((res < 0) && (res != -EWOULDBLOCK))
2595 /* Just entered TIPC_CONNECTING state; the only
2596 * difference is that return value in non-blocking
2597 * case is EINPROGRESS, rather than EALREADY.
2601 case TIPC_CONNECTING:
2603 if (previous == TIPC_CONNECTING)
2607 timeout = msecs_to_jiffies(timeout);
2608 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2609 res = tipc_wait_for_connect(sock, &timeout);
2611 case TIPC_ESTABLISHED:
2624 * tipc_listen - allow socket to listen for incoming connections
2625 * @sock: socket structure
2628 * Returns 0 on success, errno otherwise
2630 static int tipc_listen(struct socket *sock, int len)
2632 struct sock *sk = sock->sk;
2636 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2642 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2644 struct sock *sk = sock->sk;
2648 /* True wake-one mechanism for incoming connections: only
2649 * one process gets woken up, not the 'whole herd'.
2650 * Since we do not 'race & poll' for established sockets
2651 * anymore, the common case will execute the loop only once.
2654 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2655 TASK_INTERRUPTIBLE);
2656 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2658 timeo = schedule_timeout(timeo);
2662 if (!skb_queue_empty(&sk->sk_receive_queue))
2667 err = sock_intr_errno(timeo);
2668 if (signal_pending(current))
2671 finish_wait(sk_sleep(sk), &wait);
2676 * tipc_accept - wait for connection request
2677 * @sock: listening socket
2678 * @new_sock: new socket that is to be connected
2679 * @flags: file-related flags associated with socket
2681 * Returns 0 on success, errno otherwise
2683 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2686 struct sock *new_sk, *sk = sock->sk;
2687 struct sk_buff *buf;
2688 struct tipc_sock *new_tsock;
2689 struct tipc_msg *msg;
2695 if (sk->sk_state != TIPC_LISTEN) {
2699 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2700 res = tipc_wait_for_accept(sock, timeo);
2704 buf = skb_peek(&sk->sk_receive_queue);
2706 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2709 security_sk_clone(sock->sk, new_sock->sk);
2711 new_sk = new_sock->sk;
2712 new_tsock = tipc_sk(new_sk);
2715 /* we lock on new_sk; but lockdep sees the lock on sk */
2716 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2719 * Reject any stray messages received by new socket
2720 * before the socket lock was taken (very, very unlikely)
2722 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2724 /* Connect new socket to it's peer */
2725 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2727 tsk_set_importance(new_sk, msg_importance(msg));
2728 if (msg_named(msg)) {
2729 new_tsock->conn_type = msg_nametype(msg);
2730 new_tsock->conn_instance = msg_nameinst(msg);
2734 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2735 * Respond to 'SYN+' by queuing it on new socket.
2737 if (!msg_data_sz(msg)) {
2738 struct msghdr m = {NULL,};
2740 tsk_advance_rx_queue(sk);
2741 __tipc_sendstream(new_sock, &m, 0);
2743 __skb_dequeue(&sk->sk_receive_queue);
2744 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2745 skb_set_owner_r(buf, new_sk);
2747 release_sock(new_sk);
2754 * tipc_shutdown - shutdown socket connection
2755 * @sock: socket structure
2756 * @how: direction to close (must be SHUT_RDWR)
2758 * Terminates connection (if necessary), then purges socket's receive queue.
2760 * Returns 0 on success, errno otherwise
2762 static int tipc_shutdown(struct socket *sock, int how)
2764 struct sock *sk = sock->sk;
2767 if (how != SHUT_RDWR)
2772 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2773 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2774 sk->sk_shutdown = SEND_SHUTDOWN;
2776 if (sk->sk_state == TIPC_DISCONNECTING) {
2777 /* Discard any unreceived messages */
2778 __skb_queue_purge(&sk->sk_receive_queue);
2780 /* Wake up anyone sleeping in poll */
2781 sk->sk_state_change(sk);
2791 static void tipc_sk_check_probing_state(struct sock *sk,
2792 struct sk_buff_head *list)
2794 struct tipc_sock *tsk = tipc_sk(sk);
2795 u32 pnode = tsk_peer_node(tsk);
2796 u32 pport = tsk_peer_port(tsk);
2797 u32 self = tsk_own_node(tsk);
2798 u32 oport = tsk->portid;
2799 struct sk_buff *skb;
2801 if (tsk->probe_unacked) {
2802 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2803 sk->sk_err = ECONNABORTED;
2804 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2805 sk->sk_state_change(sk);
2808 /* Prepare new probe */
2809 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2810 pnode, self, pport, oport, TIPC_OK);
2812 __skb_queue_tail(list, skb);
2813 tsk->probe_unacked = true;
2814 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2817 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2819 struct tipc_sock *tsk = tipc_sk(sk);
2821 /* Try again later if dest link is congested */
2822 if (tsk->cong_link_cnt) {
2823 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2826 /* Prepare SYN for retransmit */
2827 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2830 static void tipc_sk_timeout(struct timer_list *t)
2832 struct sock *sk = from_timer(sk, t, sk_timer);
2833 struct tipc_sock *tsk = tipc_sk(sk);
2834 u32 pnode = tsk_peer_node(tsk);
2835 struct sk_buff_head list;
2838 __skb_queue_head_init(&list);
2841 /* Try again later if socket is busy */
2842 if (sock_owned_by_user(sk)) {
2843 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2849 if (sk->sk_state == TIPC_ESTABLISHED)
2850 tipc_sk_check_probing_state(sk, &list);
2851 else if (sk->sk_state == TIPC_CONNECTING)
2852 tipc_sk_retry_connect(sk, &list);
2856 if (!skb_queue_empty(&list))
2857 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2859 /* SYN messages may cause link congestion */
2860 if (rc == -ELINKCONG) {
2861 tipc_dest_push(&tsk->cong_links, pnode, 0);
2862 tsk->cong_link_cnt = 1;
2867 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2868 struct tipc_name_seq const *seq)
2870 struct sock *sk = &tsk->sk;
2871 struct net *net = sock_net(sk);
2872 struct publication *publ;
2875 if (scope != TIPC_NODE_SCOPE)
2876 scope = TIPC_CLUSTER_SCOPE;
2878 if (tipc_sk_connected(sk))
2880 key = tsk->portid + tsk->pub_count + 1;
2881 if (key == tsk->portid)
2884 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2885 scope, tsk->portid, key);
2886 if (unlikely(!publ))
2889 list_add(&publ->binding_sock, &tsk->publications);
2895 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2896 struct tipc_name_seq const *seq)
2898 struct net *net = sock_net(&tsk->sk);
2899 struct publication *publ;
2900 struct publication *safe;
2903 if (scope != TIPC_NODE_SCOPE)
2904 scope = TIPC_CLUSTER_SCOPE;
2906 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2908 if (publ->scope != scope)
2910 if (publ->type != seq->type)
2912 if (publ->lower != seq->lower)
2914 if (publ->upper != seq->upper)
2916 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2917 publ->upper, publ->key);
2921 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2922 publ->upper, publ->key);
2925 if (list_empty(&tsk->publications))
2930 /* tipc_sk_reinit: set non-zero address in all existing sockets
2931 * when we go from standalone to network mode.
2933 void tipc_sk_reinit(struct net *net)
2935 struct tipc_net *tn = net_generic(net, tipc_net_id);
2936 struct rhashtable_iter iter;
2937 struct tipc_sock *tsk;
2938 struct tipc_msg *msg;
2940 rhashtable_walk_enter(&tn->sk_rht, &iter);
2943 rhashtable_walk_start(&iter);
2945 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2946 sock_hold(&tsk->sk);
2947 rhashtable_walk_stop(&iter);
2948 lock_sock(&tsk->sk);
2950 msg_set_prevnode(msg, tipc_own_addr(net));
2951 msg_set_orignode(msg, tipc_own_addr(net));
2952 release_sock(&tsk->sk);
2953 rhashtable_walk_start(&iter);
2957 rhashtable_walk_stop(&iter);
2958 } while (tsk == ERR_PTR(-EAGAIN));
2960 rhashtable_walk_exit(&iter);
2963 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2965 struct tipc_net *tn = net_generic(net, tipc_net_id);
2966 struct tipc_sock *tsk;
2969 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2971 sock_hold(&tsk->sk);
2977 static int tipc_sk_insert(struct tipc_sock *tsk)
2979 struct sock *sk = &tsk->sk;
2980 struct net *net = sock_net(sk);
2981 struct tipc_net *tn = net_generic(net, tipc_net_id);
2982 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2983 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2985 while (remaining--) {
2987 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2988 portid = TIPC_MIN_PORT;
2989 tsk->portid = portid;
2990 sock_hold(&tsk->sk);
2991 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3000 static void tipc_sk_remove(struct tipc_sock *tsk)
3002 struct sock *sk = &tsk->sk;
3003 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3005 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3006 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3011 static const struct rhashtable_params tsk_rht_params = {
3013 .head_offset = offsetof(struct tipc_sock, node),
3014 .key_offset = offsetof(struct tipc_sock, portid),
3015 .key_len = sizeof(u32), /* portid */
3016 .max_size = 1048576,
3018 .automatic_shrinking = true,
3021 int tipc_sk_rht_init(struct net *net)
3023 struct tipc_net *tn = net_generic(net, tipc_net_id);
3025 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3028 void tipc_sk_rht_destroy(struct net *net)
3030 struct tipc_net *tn = net_generic(net, tipc_net_id);
3032 /* Wait for socket readers to complete */
3035 rhashtable_destroy(&tn->sk_rht);
3038 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3040 struct net *net = sock_net(&tsk->sk);
3041 struct tipc_group *grp = tsk->group;
3042 struct tipc_msg *hdr = &tsk->phdr;
3043 struct tipc_name_seq seq;
3046 if (mreq->type < TIPC_RESERVED_TYPES)
3048 if (mreq->scope > TIPC_NODE_SCOPE)
3052 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3056 msg_set_lookup_scope(hdr, mreq->scope);
3057 msg_set_nametype(hdr, mreq->type);
3058 msg_set_dest_droppable(hdr, true);
3059 seq.type = mreq->type;
3060 seq.lower = mreq->instance;
3061 seq.upper = seq.lower;
3062 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
3063 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
3065 tipc_group_delete(net, grp);
3069 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3070 tsk->mc_method.rcast = true;
3071 tsk->mc_method.mandatory = true;
3072 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3076 static int tipc_sk_leave(struct tipc_sock *tsk)
3078 struct net *net = sock_net(&tsk->sk);
3079 struct tipc_group *grp = tsk->group;
3080 struct tipc_name_seq seq;
3085 tipc_group_self(grp, &seq, &scope);
3086 tipc_group_delete(net, grp);
3088 tipc_sk_withdraw(tsk, scope, &seq);
3093 * tipc_setsockopt - set socket option
3094 * @sock: socket structure
3095 * @lvl: option level
3096 * @opt: option identifier
3097 * @ov: pointer to new option value
3098 * @ol: length of option value
3100 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3101 * (to ease compatibility).
3103 * Returns 0 on success, errno otherwise
3105 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3106 sockptr_t ov, unsigned int ol)
3108 struct sock *sk = sock->sk;
3109 struct tipc_sock *tsk = tipc_sk(sk);
3110 struct tipc_group_req mreq;
3114 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3116 if (lvl != SOL_TIPC)
3117 return -ENOPROTOOPT;
3120 case TIPC_IMPORTANCE:
3121 case TIPC_SRC_DROPPABLE:
3122 case TIPC_DEST_DROPPABLE:
3123 case TIPC_CONN_TIMEOUT:
3125 if (ol < sizeof(value))
3127 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3130 case TIPC_GROUP_JOIN:
3131 if (ol < sizeof(mreq))
3133 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3137 if (!sockptr_is_null(ov) || ol)
3144 case TIPC_IMPORTANCE:
3145 res = tsk_set_importance(sk, value);
3147 case TIPC_SRC_DROPPABLE:
3148 if (sock->type != SOCK_STREAM)
3149 tsk_set_unreliable(tsk, value);
3153 case TIPC_DEST_DROPPABLE:
3154 tsk_set_unreturnable(tsk, value);
3156 case TIPC_CONN_TIMEOUT:
3157 tipc_sk(sk)->conn_timeout = value;
3159 case TIPC_MCAST_BROADCAST:
3160 tsk->mc_method.rcast = false;
3161 tsk->mc_method.mandatory = true;
3163 case TIPC_MCAST_REPLICAST:
3164 tsk->mc_method.rcast = true;
3165 tsk->mc_method.mandatory = true;
3167 case TIPC_GROUP_JOIN:
3168 res = tipc_sk_join(tsk, &mreq);
3170 case TIPC_GROUP_LEAVE:
3171 res = tipc_sk_leave(tsk);
3174 tsk->nodelay = !!value;
3187 * tipc_getsockopt - get socket option
3188 * @sock: socket structure
3189 * @lvl: option level
3190 * @opt: option identifier
3191 * @ov: receptacle for option value
3192 * @ol: receptacle for length of option value
3194 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3195 * (to ease compatibility).
3197 * Returns 0 on success, errno otherwise
3199 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3200 char __user *ov, int __user *ol)
3202 struct sock *sk = sock->sk;
3203 struct tipc_sock *tsk = tipc_sk(sk);
3204 struct tipc_name_seq seq;
3209 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3210 return put_user(0, ol);
3211 if (lvl != SOL_TIPC)
3212 return -ENOPROTOOPT;
3213 res = get_user(len, ol);
3220 case TIPC_IMPORTANCE:
3221 value = tsk_importance(tsk);
3223 case TIPC_SRC_DROPPABLE:
3224 value = tsk_unreliable(tsk);
3226 case TIPC_DEST_DROPPABLE:
3227 value = tsk_unreturnable(tsk);
3229 case TIPC_CONN_TIMEOUT:
3230 value = tsk->conn_timeout;
3231 /* no need to set "res", since already 0 at this point */
3233 case TIPC_NODE_RECVQ_DEPTH:
3234 value = 0; /* was tipc_queue_size, now obsolete */
3236 case TIPC_SOCK_RECVQ_DEPTH:
3237 value = skb_queue_len(&sk->sk_receive_queue);
3239 case TIPC_SOCK_RECVQ_USED:
3240 value = sk_rmem_alloc_get(sk);
3242 case TIPC_GROUP_JOIN:
3245 tipc_group_self(tsk->group, &seq, &scope);
3255 return res; /* "get" failed */
3257 if (len < sizeof(value))
3260 if (copy_to_user(ov, &value, sizeof(value)))
3263 return put_user(sizeof(value), ol);
3266 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3268 struct net *net = sock_net(sock->sk);
3269 struct tipc_sioc_nodeid_req nr = {0};
3270 struct tipc_sioc_ln_req lnr;
3271 void __user *argp = (void __user *)arg;
3274 case SIOCGETLINKNAME:
3275 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3277 if (!tipc_node_get_linkname(net,
3278 lnr.bearer_id & 0xffff, lnr.peer,
3279 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3280 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3284 return -EADDRNOTAVAIL;
3286 if (copy_from_user(&nr, argp, sizeof(nr)))
3288 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3289 return -EADDRNOTAVAIL;
3290 if (copy_to_user(argp, &nr, sizeof(nr)))
3294 return -ENOIOCTLCMD;
3298 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3300 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3301 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3302 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3304 tsk1->peer.family = AF_TIPC;
3305 tsk1->peer.addrtype = TIPC_ADDR_ID;
3306 tsk1->peer.scope = TIPC_NODE_SCOPE;
3307 tsk1->peer.addr.id.ref = tsk2->portid;
3308 tsk1->peer.addr.id.node = onode;
3309 tsk2->peer.family = AF_TIPC;
3310 tsk2->peer.addrtype = TIPC_ADDR_ID;
3311 tsk2->peer.scope = TIPC_NODE_SCOPE;
3312 tsk2->peer.addr.id.ref = tsk1->portid;
3313 tsk2->peer.addr.id.node = onode;
3315 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3316 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3320 /* Protocol switches for the various types of TIPC sockets */
3322 static const struct proto_ops msg_ops = {
3323 .owner = THIS_MODULE,
3325 .release = tipc_release,
3327 .connect = tipc_connect,
3328 .socketpair = tipc_socketpair,
3329 .accept = sock_no_accept,
3330 .getname = tipc_getname,
3332 .ioctl = tipc_ioctl,
3333 .listen = sock_no_listen,
3334 .shutdown = tipc_shutdown,
3335 .setsockopt = tipc_setsockopt,
3336 .getsockopt = tipc_getsockopt,
3337 .sendmsg = tipc_sendmsg,
3338 .recvmsg = tipc_recvmsg,
3339 .mmap = sock_no_mmap,
3340 .sendpage = sock_no_sendpage
3343 static const struct proto_ops packet_ops = {
3344 .owner = THIS_MODULE,
3346 .release = tipc_release,
3348 .connect = tipc_connect,
3349 .socketpair = tipc_socketpair,
3350 .accept = tipc_accept,
3351 .getname = tipc_getname,
3353 .ioctl = tipc_ioctl,
3354 .listen = tipc_listen,
3355 .shutdown = tipc_shutdown,
3356 .setsockopt = tipc_setsockopt,
3357 .getsockopt = tipc_getsockopt,
3358 .sendmsg = tipc_send_packet,
3359 .recvmsg = tipc_recvmsg,
3360 .mmap = sock_no_mmap,
3361 .sendpage = sock_no_sendpage
3364 static const struct proto_ops stream_ops = {
3365 .owner = THIS_MODULE,
3367 .release = tipc_release,
3369 .connect = tipc_connect,
3370 .socketpair = tipc_socketpair,
3371 .accept = tipc_accept,
3372 .getname = tipc_getname,
3374 .ioctl = tipc_ioctl,
3375 .listen = tipc_listen,
3376 .shutdown = tipc_shutdown,
3377 .setsockopt = tipc_setsockopt,
3378 .getsockopt = tipc_getsockopt,
3379 .sendmsg = tipc_sendstream,
3380 .recvmsg = tipc_recvstream,
3381 .mmap = sock_no_mmap,
3382 .sendpage = sock_no_sendpage
3385 static const struct net_proto_family tipc_family_ops = {
3386 .owner = THIS_MODULE,
3388 .create = tipc_sk_create
3391 static struct proto tipc_proto = {
3393 .owner = THIS_MODULE,
3394 .obj_size = sizeof(struct tipc_sock),
3395 .sysctl_rmem = sysctl_tipc_rmem
3399 * tipc_socket_init - initialize TIPC socket interface
3401 * Returns 0 on success, errno otherwise
3403 int tipc_socket_init(void)
3407 res = proto_register(&tipc_proto, 1);
3409 pr_err("Failed to register TIPC protocol type\n");
3413 res = sock_register(&tipc_family_ops);
3415 pr_err("Failed to register TIPC socket type\n");
3416 proto_unregister(&tipc_proto);
3424 * tipc_socket_stop - stop TIPC socket interface
3426 void tipc_socket_stop(void)
3428 sock_unregister(tipc_family_ops.family);
3429 proto_unregister(&tipc_proto);
3432 /* Caller should hold socket lock for the passed tipc socket. */
3433 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3437 struct nlattr *nest;
3439 peer_node = tsk_peer_node(tsk);
3440 peer_port = tsk_peer_port(tsk);
3442 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3446 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3448 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3451 if (tsk->conn_type != 0) {
3452 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3454 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3456 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3459 nla_nest_end(skb, nest);
3464 nla_nest_cancel(skb, nest);
3469 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3472 struct net *net = sock_net(skb->sk);
3473 struct sock *sk = &tsk->sk;
3475 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3476 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3479 if (tipc_sk_connected(sk)) {
3480 if (__tipc_nl_add_sk_con(skb, tsk))
3482 } else if (!list_empty(&tsk->publications)) {
3483 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3489 /* Caller should hold socket lock for the passed tipc socket. */
3490 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3491 struct tipc_sock *tsk)
3493 struct nlattr *attrs;
3496 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3497 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3501 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3503 goto genlmsg_cancel;
3505 if (__tipc_nl_add_sk_info(skb, tsk))
3506 goto attr_msg_cancel;
3508 nla_nest_end(skb, attrs);
3509 genlmsg_end(skb, hdr);
3514 nla_nest_cancel(skb, attrs);
3516 genlmsg_cancel(skb, hdr);
3521 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3522 int (*skb_handler)(struct sk_buff *skb,
3523 struct netlink_callback *cb,
3524 struct tipc_sock *tsk))
3526 struct rhashtable_iter *iter = (void *)cb->args[4];
3527 struct tipc_sock *tsk;
3530 rhashtable_walk_start(iter);
3531 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3534 if (err == -EAGAIN) {
3541 sock_hold(&tsk->sk);
3542 rhashtable_walk_stop(iter);
3543 lock_sock(&tsk->sk);
3544 err = skb_handler(skb, cb, tsk);
3546 release_sock(&tsk->sk);
3550 release_sock(&tsk->sk);
3551 rhashtable_walk_start(iter);
3554 rhashtable_walk_stop(iter);
3558 EXPORT_SYMBOL(tipc_nl_sk_walk);
3560 int tipc_dump_start(struct netlink_callback *cb)
3562 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3564 EXPORT_SYMBOL(tipc_dump_start);
3566 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3568 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3569 struct rhashtable_iter *iter = (void *)cb->args[4];
3570 struct tipc_net *tn = tipc_net(net);
3573 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3577 cb->args[4] = (long)iter;
3580 rhashtable_walk_enter(&tn->sk_rht, iter);
3584 int tipc_dump_done(struct netlink_callback *cb)
3586 struct rhashtable_iter *hti = (void *)cb->args[4];
3588 rhashtable_walk_exit(hti);
3592 EXPORT_SYMBOL(tipc_dump_done);
3594 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3595 struct tipc_sock *tsk, u32 sk_filter_state,
3596 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3598 struct sock *sk = &tsk->sk;
3599 struct nlattr *attrs;
3600 struct nlattr *stat;
3602 /*filter response w.r.t sk_state*/
3603 if (!(sk_filter_state & (1 << sk->sk_state)))
3606 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3610 if (__tipc_nl_add_sk_info(skb, tsk))
3611 goto attr_msg_cancel;
3613 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3614 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3615 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3616 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3617 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3619 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3620 tipc_diag_gen_cookie(sk),
3622 goto attr_msg_cancel;
3624 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3626 goto attr_msg_cancel;
3628 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3629 skb_queue_len(&sk->sk_receive_queue)) ||
3630 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3631 skb_queue_len(&sk->sk_write_queue)) ||
3632 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3633 atomic_read(&sk->sk_drops)))
3634 goto stat_msg_cancel;
3636 if (tsk->cong_link_cnt &&
3637 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3638 goto stat_msg_cancel;
3640 if (tsk_conn_cong(tsk) &&
3641 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3642 goto stat_msg_cancel;
3644 nla_nest_end(skb, stat);
3647 if (tipc_group_fill_sock_diag(tsk->group, skb))
3648 goto stat_msg_cancel;
3650 nla_nest_end(skb, attrs);
3655 nla_nest_cancel(skb, stat);
3657 nla_nest_cancel(skb, attrs);
3661 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3663 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3665 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3668 /* Caller should hold socket lock for the passed tipc socket. */
3669 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3670 struct netlink_callback *cb,
3671 struct publication *publ)
3674 struct nlattr *attrs;
3676 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3677 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3681 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3683 goto genlmsg_cancel;
3685 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3686 goto attr_msg_cancel;
3687 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3688 goto attr_msg_cancel;
3689 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3690 goto attr_msg_cancel;
3691 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3692 goto attr_msg_cancel;
3694 nla_nest_end(skb, attrs);
3695 genlmsg_end(skb, hdr);
3700 nla_nest_cancel(skb, attrs);
3702 genlmsg_cancel(skb, hdr);
3707 /* Caller should hold socket lock for the passed tipc socket. */
3708 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3709 struct netlink_callback *cb,
3710 struct tipc_sock *tsk, u32 *last_publ)
3713 struct publication *p;
3716 list_for_each_entry(p, &tsk->publications, binding_sock) {
3717 if (p->key == *last_publ)
3720 if (p->key != *last_publ) {
3721 /* We never set seq or call nl_dump_check_consistent()
3722 * this means that setting prev_seq here will cause the
3723 * consistence check to fail in the netlink callback
3724 * handler. Resulting in the last NLMSG_DONE message
3725 * having the NLM_F_DUMP_INTR flag set.
3732 p = list_first_entry(&tsk->publications, struct publication,
3736 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3737 err = __tipc_nl_add_sk_publ(skb, cb, p);
3739 *last_publ = p->key;
3748 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3751 u32 tsk_portid = cb->args[0];
3752 u32 last_publ = cb->args[1];
3753 u32 done = cb->args[2];
3754 struct net *net = sock_net(skb->sk);
3755 struct tipc_sock *tsk;
3758 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3759 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3761 if (!attrs[TIPC_NLA_SOCK])
3764 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3765 attrs[TIPC_NLA_SOCK],
3766 tipc_nl_sock_policy, NULL);
3770 if (!sock[TIPC_NLA_SOCK_REF])
3773 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3779 tsk = tipc_sk_lookup(net, tsk_portid);
3783 lock_sock(&tsk->sk);
3784 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3787 release_sock(&tsk->sk);
3790 cb->args[0] = tsk_portid;
3791 cb->args[1] = last_publ;
3798 * tipc_sk_filtering - check if a socket should be traced
3799 * @sk: the socket to be examined
3800 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3801 * (portid, sock type, name type, name lower, name upper)
3803 * Returns true if the socket meets the socket tuple data
3804 * (value 0 = 'any') or when there is no tuple set (all = 0),
3807 bool tipc_sk_filtering(struct sock *sk)
3809 struct tipc_sock *tsk;
3810 struct publication *p;
3811 u32 _port, _sktype, _type, _lower, _upper;
3812 u32 type = 0, lower = 0, upper = 0;
3819 _port = sysctl_tipc_sk_filter[0];
3820 _sktype = sysctl_tipc_sk_filter[1];
3821 _type = sysctl_tipc_sk_filter[2];
3822 _lower = sysctl_tipc_sk_filter[3];
3823 _upper = sysctl_tipc_sk_filter[4];
3825 if (!_port && !_sktype && !_type && !_lower && !_upper)
3829 return (_port == tsk->portid);
3831 if (_sktype && _sktype != sk->sk_type)
3834 if (tsk->published) {
3835 p = list_first_entry_or_null(&tsk->publications,
3836 struct publication, binding_sock);
3844 if (!tipc_sk_type_connectionless(sk)) {
3845 type = tsk->conn_type;
3846 lower = tsk->conn_instance;
3847 upper = tsk->conn_instance;
3850 if ((_type && _type != type) || (_lower && _lower != lower) ||
3851 (_upper && _upper != upper))
3857 u32 tipc_sock_get_portid(struct sock *sk)
3859 return (sk) ? (tipc_sk(sk))->portid : 0;
3863 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3864 * both the rcv and backlog queues are considered
3865 * @sk: tipc sk to be checked
3866 * @skb: tipc msg to be checked
3868 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3871 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3873 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3874 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3875 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3877 return (qsize > lim * 90 / 100);
3881 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3882 * only the rcv queue is considered
3883 * @sk: tipc sk to be checked
3884 * @skb: tipc msg to be checked
3886 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3889 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3891 unsigned int lim = rcvbuf_limit(sk, skb);
3892 unsigned int qsize = sk_rmem_alloc_get(sk);
3894 return (qsize > lim * 90 / 100);
3898 * tipc_sk_dump - dump TIPC socket
3899 * @sk: tipc sk to be dumped
3900 * @dqueues: bitmask to decide if any socket queue to be dumped?
3901 * - TIPC_DUMP_NONE: don't dump socket queues
3902 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3903 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3904 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3905 * - TIPC_DUMP_ALL: dump all the socket queues above
3906 * @buf: returned buffer of dump data in format
3908 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3911 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3912 struct tipc_sock *tsk;
3913 struct publication *p;
3917 i += scnprintf(buf, sz, "sk data: (null)\n");
3922 tsk_connected = !tipc_sk_type_connectionless(sk);
3924 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3925 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3926 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3927 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3928 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3929 if (tsk_connected) {
3930 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3931 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3932 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3933 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3935 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3936 if (tsk->published) {
3937 p = list_first_entry_or_null(&tsk->publications,
3938 struct publication, binding_sock);
3939 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3940 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3941 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3943 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3944 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3945 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3946 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3947 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3948 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3949 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3950 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3951 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3952 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3953 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3954 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3955 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3956 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3958 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3959 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3960 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3963 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3964 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3965 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3968 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3969 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3970 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3971 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3972 i += scnprintf(buf + i, sz - i, " tail ");
3973 i += tipc_skb_dump(sk->sk_backlog.tail, false,