2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * Copyright (c) 2020-2021, Red Hat Inc
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the names of the copyright holders nor the names of its
18 * contributors may be used to endorse or promote products derived from
19 * this software without specific prior written permission.
21 * Alternatively, this software may be distributed under the terms of the
22 * GNU General Public License ("GPL") version 2 as published by the Free
23 * Software Foundation.
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
38 #include <linux/rhashtable.h>
39 #include <linux/sched/signal.h>
42 #include "name_table.h"
45 #include "name_distr.h"
52 #define NAGLE_START_INIT 4
53 #define NAGLE_START_MAX 1024
54 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
55 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
56 #define TIPC_MAX_PORT 0xffffffff
57 #define TIPC_MIN_PORT 1
58 #define TIPC_ACK_RATE 4 /* ACK at 1/4 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 * @probe_unacked: probe has not received ack yet
89 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
90 * @cong_link_cnt: number of congested links
91 * @snt_unacked: # messages sent by socket, and not yet acked by peer
92 * @snd_win: send window size
93 * @peer_caps: peer capabilities mask
94 * @rcv_unacked: # messages read by user, but not yet acked back to peer
95 * @rcv_win: receive window size
96 * @peer: 'connected' peer for dgram/rdm
97 * @node: hash table node
98 * @mc_method: cookie for use between socket and broadcast layer
99 * @rcu: rcu struct for tipc_sock
100 * @group: TIPC communications group
101 * @oneway: message count in one direction (FIXME)
102 * @nagle_start: current nagle value
103 * @snd_backlog: send backlog count
104 * @msg_acc: messages accepted; used in managing backlog and nagle
105 * @pkt_cnt: TIPC socket packet count
106 * @expect_ack: whether this TIPC socket is expecting an ack
107 * @nodelay: setsockopt() TIPC_NODELAY setting
108 * @group_is_open: TIPC socket group is fully open (FIXME)
117 struct tipc_msg phdr;
118 struct list_head cong_links;
119 struct list_head publications;
121 atomic_t dupl_rcvcnt;
130 struct sockaddr_tipc peer;
131 struct rhash_head node;
132 struct tipc_mc_method mc_method;
134 struct tipc_group *group;
146 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
147 static void tipc_data_ready(struct sock *sk);
148 static void tipc_write_space(struct sock *sk);
149 static void tipc_sock_destruct(struct sock *sk);
150 static int tipc_release(struct socket *sock);
151 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
153 static void tipc_sk_timeout(struct timer_list *t);
154 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
155 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
156 static int tipc_sk_leave(struct tipc_sock *tsk);
157 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
158 static int tipc_sk_insert(struct tipc_sock *tsk);
159 static void tipc_sk_remove(struct tipc_sock *tsk);
160 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
161 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
162 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
163 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
165 static const struct proto_ops packet_ops;
166 static const struct proto_ops stream_ops;
167 static const struct proto_ops msg_ops;
168 static struct proto tipc_proto;
169 static const struct rhashtable_params tsk_rht_params;
171 static u32 tsk_own_node(struct tipc_sock *tsk)
173 return msg_prevnode(&tsk->phdr);
176 static u32 tsk_peer_node(struct tipc_sock *tsk)
178 return msg_destnode(&tsk->phdr);
181 static u32 tsk_peer_port(struct tipc_sock *tsk)
183 return msg_destport(&tsk->phdr);
186 static bool tsk_unreliable(struct tipc_sock *tsk)
188 return msg_src_droppable(&tsk->phdr) != 0;
191 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
193 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
196 static bool tsk_unreturnable(struct tipc_sock *tsk)
198 return msg_dest_droppable(&tsk->phdr) != 0;
201 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
203 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
206 static int tsk_importance(struct tipc_sock *tsk)
208 return msg_importance(&tsk->phdr);
211 static struct tipc_sock *tipc_sk(const struct sock *sk)
213 return container_of(sk, struct tipc_sock, sk);
216 int tsk_set_importance(struct sock *sk, int imp)
218 if (imp > TIPC_CRITICAL_IMPORTANCE)
220 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
224 static bool tsk_conn_cong(struct tipc_sock *tsk)
226 return tsk->snt_unacked > tsk->snd_win;
229 static u16 tsk_blocks(int len)
231 return ((len / FLOWCTL_BLK_SZ) + 1);
234 /* tsk_blocks(): translate a buffer size in bytes to number of
235 * advertisable blocks, taking into account the ratio truesize(len)/len
236 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
238 static u16 tsk_adv_blocks(int len)
240 return len / FLOWCTL_BLK_SZ / 4;
243 /* tsk_inc(): increment counter for sent or received data
244 * - If block based flow control is not supported by peer we
245 * fall back to message based ditto, incrementing the counter
247 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
249 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
250 return ((msglen / FLOWCTL_BLK_SZ) + 1);
254 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
256 static void tsk_set_nagle(struct tipc_sock *tsk)
258 struct sock *sk = &tsk->sk;
261 if (sk->sk_type != SOCK_STREAM)
265 if (!(tsk->peer_caps & TIPC_NAGLE))
267 /* Limit node local buffer size to avoid receive queue overflow */
268 if (tsk->max_pkt == MAX_MSG_SIZE)
269 tsk->maxnagle = 1500;
271 tsk->maxnagle = tsk->max_pkt;
275 * tsk_advance_rx_queue - discard first buffer in socket receive queue
276 * @sk: network socket
278 * Caller must hold socket lock
280 static void tsk_advance_rx_queue(struct sock *sk)
282 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
283 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
286 /* tipc_sk_respond() : send response message back to sender
288 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
292 u32 onode = tipc_own_addr(sock_net(sk));
294 if (!tipc_msg_reverse(onode, &skb, err))
297 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
298 dnode = msg_destnode(buf_msg(skb));
299 selector = msg_origport(buf_msg(skb));
300 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
304 * tsk_rej_rx_queue - reject all buffers in socket receive queue
305 * @sk: network socket
306 * @error: response error code
308 * Caller must hold socket lock
310 static void tsk_rej_rx_queue(struct sock *sk, int error)
314 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
315 tipc_sk_respond(sk, skb, error);
318 static bool tipc_sk_connected(struct sock *sk)
320 return sk->sk_state == TIPC_ESTABLISHED;
323 /* tipc_sk_type_connectionless - check if the socket is datagram socket
326 * Returns true if connection less, false otherwise
328 static bool tipc_sk_type_connectionless(struct sock *sk)
330 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
333 /* tsk_peer_msg - verify if message was sent by connected port's peer
335 * Handles cases where the node's network address has changed from
336 * the default of <0.0.0> to its configured setting.
338 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
340 struct sock *sk = &tsk->sk;
341 u32 self = tipc_own_addr(sock_net(sk));
342 u32 peer_port = tsk_peer_port(tsk);
343 u32 orig_node, peer_node;
345 if (unlikely(!tipc_sk_connected(sk)))
348 if (unlikely(msg_origport(msg) != peer_port))
351 orig_node = msg_orignode(msg);
352 peer_node = tsk_peer_node(tsk);
354 if (likely(orig_node == peer_node))
357 if (!orig_node && peer_node == self)
360 if (!peer_node && orig_node == self)
366 /* tipc_set_sk_state - set the sk_state of the socket
369 * Caller must hold socket lock
371 * Returns 0 on success, errno otherwise
373 static int tipc_set_sk_state(struct sock *sk, int state)
375 int oldsk_state = sk->sk_state;
383 case TIPC_CONNECTING:
384 if (oldsk_state == TIPC_OPEN)
387 case TIPC_ESTABLISHED:
388 if (oldsk_state == TIPC_CONNECTING ||
389 oldsk_state == TIPC_OPEN)
392 case TIPC_DISCONNECTING:
393 if (oldsk_state == TIPC_CONNECTING ||
394 oldsk_state == TIPC_ESTABLISHED)
400 sk->sk_state = state;
405 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
407 struct sock *sk = sock->sk;
408 int err = sock_error(sk);
409 int typ = sock->type;
413 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
414 if (sk->sk_state == TIPC_DISCONNECTING)
416 else if (!tipc_sk_connected(sk))
421 if (signal_pending(current))
422 return sock_intr_errno(*timeout);
427 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
429 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
433 while ((rc_ = !(condition_))) { \
434 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
437 rc_ = tipc_sk_sock_err((sock_), timeo_); \
440 add_wait_queue(sk_sleep(sk_), &wait_); \
442 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
443 sched_annotate_sleep(); \
445 remove_wait_queue(sk_sleep(sk_), &wait_); \
451 * tipc_sk_create - create a TIPC socket
452 * @net: network namespace (must be default network)
453 * @sock: pre-allocated socket structure
454 * @protocol: protocol indicator (must be 0)
455 * @kern: caused by kernel or by userspace?
457 * This routine creates additional data structures used by the TIPC socket,
458 * initializes them, and links them together.
460 * Return: 0 on success, errno otherwise
462 static int tipc_sk_create(struct net *net, struct socket *sock,
463 int protocol, int kern)
465 const struct proto_ops *ops;
467 struct tipc_sock *tsk;
468 struct tipc_msg *msg;
470 /* Validate arguments */
471 if (unlikely(protocol != 0))
472 return -EPROTONOSUPPORT;
474 switch (sock->type) {
489 /* Allocate socket's protocol area */
490 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
495 tsk->max_pkt = MAX_PKT_DEFAULT;
497 tsk->nagle_start = NAGLE_START_INIT;
498 INIT_LIST_HEAD(&tsk->publications);
499 INIT_LIST_HEAD(&tsk->cong_links);
502 /* Finish initializing socket data structures */
504 sock_init_data(sock, sk);
505 tipc_set_sk_state(sk, TIPC_OPEN);
506 if (tipc_sk_insert(tsk)) {
507 pr_warn("Socket create failed; port number exhausted\n");
511 /* Ensure tsk is visible before we read own_addr. */
514 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
515 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
517 msg_set_origport(msg, tsk->portid);
518 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
520 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
521 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
522 sk->sk_data_ready = tipc_data_ready;
523 sk->sk_write_space = tipc_write_space;
524 sk->sk_destruct = tipc_sock_destruct;
525 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
526 tsk->group_is_open = true;
527 atomic_set(&tsk->dupl_rcvcnt, 0);
529 /* Start out with safe limits until we receive an advertised window */
530 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
531 tsk->rcv_win = tsk->snd_win;
533 if (tipc_sk_type_connectionless(sk)) {
534 tsk_set_unreturnable(tsk, true);
535 if (sock->type == SOCK_DGRAM)
536 tsk_set_unreliable(tsk, true);
538 __skb_queue_head_init(&tsk->mc_method.deferredq);
539 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
543 static void tipc_sk_callback(struct rcu_head *head)
545 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
550 /* Caller should hold socket lock for the socket. */
551 static void __tipc_shutdown(struct socket *sock, int error)
553 struct sock *sk = sock->sk;
554 struct tipc_sock *tsk = tipc_sk(sk);
555 struct net *net = sock_net(sk);
556 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
557 u32 dnode = tsk_peer_node(tsk);
560 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
561 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
562 !tsk_conn_cong(tsk)));
564 /* Push out delayed messages if in Nagle mode */
565 tipc_sk_push_backlog(tsk, false);
566 /* Remove pending SYN */
567 __skb_queue_purge(&sk->sk_write_queue);
569 /* Remove partially received buffer if any */
570 skb = skb_peek(&sk->sk_receive_queue);
571 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
572 __skb_unlink(skb, &sk->sk_receive_queue);
576 /* Reject all unreceived messages if connectionless */
577 if (tipc_sk_type_connectionless(sk)) {
578 tsk_rej_rx_queue(sk, error);
582 switch (sk->sk_state) {
583 case TIPC_CONNECTING:
584 case TIPC_ESTABLISHED:
585 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
586 tipc_node_remove_conn(net, dnode, tsk->portid);
587 /* Send a FIN+/- to its peer */
588 skb = __skb_dequeue(&sk->sk_receive_queue);
590 __skb_queue_purge(&sk->sk_receive_queue);
591 tipc_sk_respond(sk, skb, error);
594 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
595 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
596 tsk_own_node(tsk), tsk_peer_port(tsk),
599 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
602 /* Reject all SYN messages */
603 tsk_rej_rx_queue(sk, error);
606 __skb_queue_purge(&sk->sk_receive_queue);
612 * tipc_release - destroy a TIPC socket
613 * @sock: socket to destroy
615 * This routine cleans up any messages that are still queued on the socket.
616 * For DGRAM and RDM socket types, all queued messages are rejected.
617 * For SEQPACKET and STREAM socket types, the first message is rejected
618 * and any others are discarded. (If the first message on a STREAM socket
619 * is partially-read, it is discarded and the next one is rejected instead.)
621 * NOTE: Rejected messages are not necessarily returned to the sender! They
622 * are returned or discarded according to the "destination droppable" setting
623 * specified for the message by the sender.
625 * Return: 0 on success, errno otherwise
627 static int tipc_release(struct socket *sock)
629 struct sock *sk = sock->sk;
630 struct tipc_sock *tsk;
633 * Exit if socket isn't fully initialized (occurs when a failed accept()
634 * releases a pre-allocated child socket that was never used)
642 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
643 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
644 sk->sk_shutdown = SHUTDOWN_MASK;
646 tipc_sk_withdraw(tsk, NULL);
647 __skb_queue_purge(&tsk->mc_method.deferredq);
648 sk_stop_timer(sk, &sk->sk_timer);
652 /* Reject any messages that accumulated in backlog queue */
654 tipc_dest_list_purge(&tsk->cong_links);
655 tsk->cong_link_cnt = 0;
656 call_rcu(&tsk->rcu, tipc_sk_callback);
663 * __tipc_bind - associate or disassocate TIPC name(s) with a socket
664 * @sock: socket structure
665 * @skaddr: socket address describing name(s) and desired operation
666 * @alen: size of socket address data structure
668 * Name and name sequence binding is indicated using a positive scope value;
669 * a negative scope value unbinds the specified name. Specifying no name
670 * (i.e. a socket address length of 0) unbinds all names from the socket.
672 * Return: 0 on success, errno otherwise
674 * NOTE: This routine doesn't need to take the socket lock since it doesn't
675 * access any non-constant socket information.
677 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
679 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
680 struct tipc_sock *tsk = tipc_sk(sock->sk);
684 return tipc_sk_withdraw(tsk, NULL);
686 if (ua->addrtype == TIPC_SERVICE_ADDR) {
687 ua->addrtype = TIPC_SERVICE_RANGE;
688 ua->sr.upper = ua->sr.lower;
692 ua->scope = -ua->scope;
694 /* Users may still use deprecated TIPC_ZONE_SCOPE */
695 if (ua->scope != TIPC_NODE_SCOPE)
696 ua->scope = TIPC_CLUSTER_SCOPE;
702 return tipc_sk_withdraw(tsk, ua);
703 return tipc_sk_publish(tsk, ua);
706 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
711 res = __tipc_bind(sock, skaddr, alen);
712 release_sock(sock->sk);
716 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
718 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
719 u32 atype = ua->addrtype;
722 if (!tipc_uaddr_valid(ua, alen))
724 if (atype == TIPC_SOCKET_ADDR)
725 return -EAFNOSUPPORT;
726 if (ua->sr.type < TIPC_RESERVED_TYPES) {
727 pr_warn_once("Can't bind to reserved service type %u\n",
732 return tipc_sk_bind(sock, skaddr, alen);
736 * tipc_getname - get port ID of socket or peer socket
737 * @sock: socket structure
738 * @uaddr: area for returned socket address
739 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
741 * Return: 0 on success, errno otherwise
743 * NOTE: This routine doesn't need to take the socket lock since it only
744 * accesses socket information that is unchanging (or which changes in
745 * a completely predictable manner).
747 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
750 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
751 struct sock *sk = sock->sk;
752 struct tipc_sock *tsk = tipc_sk(sk);
754 memset(addr, 0, sizeof(*addr));
756 if ((!tipc_sk_connected(sk)) &&
757 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
759 addr->addr.id.ref = tsk_peer_port(tsk);
760 addr->addr.id.node = tsk_peer_node(tsk);
762 addr->addr.id.ref = tsk->portid;
763 addr->addr.id.node = tipc_own_addr(sock_net(sk));
766 addr->addrtype = TIPC_SOCKET_ADDR;
767 addr->family = AF_TIPC;
769 addr->addr.name.domain = 0;
771 return sizeof(*addr);
775 * tipc_poll - read and possibly block on pollmask
776 * @file: file structure associated with the socket
777 * @sock: socket for which to calculate the poll bits
780 * Return: pollmask value
783 * It appears that the usual socket locking mechanisms are not useful here
784 * since the pollmask info is potentially out-of-date the moment this routine
785 * exits. TCP and other protocols seem to rely on higher level poll routines
786 * to handle any preventable race conditions, so TIPC will do the same ...
788 * IMPORTANT: The fact that a read or write operation is indicated does NOT
789 * imply that the operation will succeed, merely that it should be performed
790 * and will not block.
792 static __poll_t tipc_poll(struct file *file, struct socket *sock,
795 struct sock *sk = sock->sk;
796 struct tipc_sock *tsk = tipc_sk(sk);
797 __poll_t revents = 0;
799 sock_poll_wait(file, sock, wait);
800 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
802 if (sk->sk_shutdown & RCV_SHUTDOWN)
803 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
804 if (sk->sk_shutdown == SHUTDOWN_MASK)
807 switch (sk->sk_state) {
808 case TIPC_ESTABLISHED:
809 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
813 case TIPC_CONNECTING:
814 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
815 revents |= EPOLLIN | EPOLLRDNORM;
818 if (tsk->group_is_open && !tsk->cong_link_cnt)
820 if (!tipc_sk_type_connectionless(sk))
822 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
824 revents |= EPOLLIN | EPOLLRDNORM;
826 case TIPC_DISCONNECTING:
827 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
834 * tipc_sendmcast - send multicast message
835 * @sock: socket structure
836 * @ua: destination address struct
837 * @msg: message to send
838 * @dlen: length of data to send
839 * @timeout: timeout to wait for wakeup
841 * Called from function tipc_sendmsg(), which has done all sanity checks
842 * Return: the number of bytes sent on success, or errno
844 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
845 struct msghdr *msg, size_t dlen, long timeout)
847 struct sock *sk = sock->sk;
848 struct tipc_sock *tsk = tipc_sk(sk);
849 struct tipc_msg *hdr = &tsk->phdr;
850 struct net *net = sock_net(sk);
851 int mtu = tipc_bcast_get_mtu(net);
852 struct sk_buff_head pkts;
853 struct tipc_nlist dsts;
859 /* Block or return if any destination link is congested */
860 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
864 /* Lookup destination nodes */
865 tipc_nlist_init(&dsts, tipc_own_addr(net));
866 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
867 if (!dsts.local && !dsts.remote)
868 return -EHOSTUNREACH;
870 /* Build message header */
871 msg_set_type(hdr, TIPC_MCAST_MSG);
872 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
873 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
874 msg_set_destport(hdr, 0);
875 msg_set_destnode(hdr, 0);
876 msg_set_nametype(hdr, ua->sr.type);
877 msg_set_namelower(hdr, ua->sr.lower);
878 msg_set_nameupper(hdr, ua->sr.upper);
880 /* Build message as chain of buffers */
881 __skb_queue_head_init(&pkts);
882 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
884 /* Send message if build was successful */
885 if (unlikely(rc == dlen)) {
886 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
887 TIPC_DUMP_SK_SNDQ, " ");
888 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
889 &tsk->cong_link_cnt);
892 tipc_nlist_purge(&dsts);
894 return rc ? rc : dlen;
898 * tipc_send_group_msg - send a message to a member in the group
899 * @net: network namespace
901 * @m: message to send
903 * @dnode: destination node
904 * @dport: destination port
905 * @dlen: total length of message data
907 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
908 struct msghdr *m, struct tipc_member *mb,
909 u32 dnode, u32 dport, int dlen)
911 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
912 struct tipc_mc_method *method = &tsk->mc_method;
913 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
914 struct tipc_msg *hdr = &tsk->phdr;
915 struct sk_buff_head pkts;
918 /* Complete message header */
919 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
920 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
921 msg_set_destport(hdr, dport);
922 msg_set_destnode(hdr, dnode);
923 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
925 /* Build message as chain of buffers */
926 __skb_queue_head_init(&pkts);
927 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
928 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
929 if (unlikely(rc != dlen))
933 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
934 if (unlikely(rc == -ELINKCONG)) {
935 tipc_dest_push(&tsk->cong_links, dnode, 0);
936 tsk->cong_link_cnt++;
939 /* Update send window */
940 tipc_group_update_member(mb, blks);
942 /* A broadcast sent within next EXPIRE period must follow same path */
943 method->rcast = true;
944 method->mandatory = true;
949 * tipc_send_group_unicast - send message to a member in the group
950 * @sock: socket structure
951 * @m: message to send
952 * @dlen: total length of message data
953 * @timeout: timeout to wait for wakeup
955 * Called from function tipc_sendmsg(), which has done all sanity checks
956 * Return: the number of bytes sent on success, or errno
958 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
959 int dlen, long timeout)
961 struct sock *sk = sock->sk;
962 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
963 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
964 struct tipc_sock *tsk = tipc_sk(sk);
965 struct net *net = sock_net(sk);
966 struct tipc_member *mb = NULL;
973 return -EHOSTUNREACH;
975 /* Block or return if destination link or member is congested */
976 rc = tipc_wait_for_cond(sock, &timeout,
977 !tipc_dest_find(&tsk->cong_links, node, 0) &&
979 !tipc_group_cong(tsk->group, node, port, blks,
985 return -EHOSTUNREACH;
987 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
989 return rc ? rc : dlen;
993 * tipc_send_group_anycast - send message to any member with given identity
994 * @sock: socket structure
995 * @m: message to send
996 * @dlen: total length of message data
997 * @timeout: timeout to wait for wakeup
999 * Called from function tipc_sendmsg(), which has done all sanity checks
1000 * Return: the number of bytes sent on success, or errno
1002 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1003 int dlen, long timeout)
1005 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1006 struct sock *sk = sock->sk;
1007 struct tipc_sock *tsk = tipc_sk(sk);
1008 struct list_head *cong_links = &tsk->cong_links;
1009 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1010 struct tipc_msg *hdr = &tsk->phdr;
1011 struct tipc_member *first = NULL;
1012 struct tipc_member *mbr = NULL;
1013 struct net *net = sock_net(sk);
1014 u32 node, port, exclude;
1015 struct list_head dsts;
1020 INIT_LIST_HEAD(&dsts);
1021 ua->sa.type = msg_nametype(hdr);
1022 ua->scope = msg_lookup_scope(hdr);
1024 while (++lookups < 4) {
1025 exclude = tipc_group_exclude(tsk->group);
1029 /* Look for a non-congested destination member, if any */
1031 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1033 return -EHOSTUNREACH;
1034 tipc_dest_pop(&dsts, &node, &port);
1035 cong = tipc_group_cong(tsk->group, node, port, blks,
1045 /* Start over if destination was not in member list */
1049 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1052 /* Block or return if destination link or member is congested */
1053 rc = tipc_wait_for_cond(sock, &timeout,
1054 !tipc_dest_find(cong_links, node, 0) &&
1056 !tipc_group_cong(tsk->group, node, port,
1061 /* Send, unless destination disappeared while waiting */
1066 if (unlikely(lookups >= 4))
1067 return -EHOSTUNREACH;
1069 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1071 return rc ? rc : dlen;
1075 * tipc_send_group_bcast - send message to all members in communication group
1076 * @sock: socket structure
1077 * @m: message to send
1078 * @dlen: total length of message data
1079 * @timeout: timeout to wait for wakeup
1081 * Called from function tipc_sendmsg(), which has done all sanity checks
1082 * Return: the number of bytes sent on success, or errno
1084 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1085 int dlen, long timeout)
1087 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1088 struct sock *sk = sock->sk;
1089 struct net *net = sock_net(sk);
1090 struct tipc_sock *tsk = tipc_sk(sk);
1091 struct tipc_nlist *dsts;
1092 struct tipc_mc_method *method = &tsk->mc_method;
1093 bool ack = method->mandatory && method->rcast;
1094 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1095 struct tipc_msg *hdr = &tsk->phdr;
1096 int mtu = tipc_bcast_get_mtu(net);
1097 struct sk_buff_head pkts;
1098 int rc = -EHOSTUNREACH;
1100 /* Block or return if any destination link or member is congested */
1101 rc = tipc_wait_for_cond(sock, &timeout,
1102 !tsk->cong_link_cnt && tsk->group &&
1103 !tipc_group_bc_cong(tsk->group, blks));
1107 dsts = tipc_group_dests(tsk->group);
1108 if (!dsts->local && !dsts->remote)
1109 return -EHOSTUNREACH;
1111 /* Complete message header */
1113 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1114 msg_set_nameinst(hdr, ua->sa.instance);
1116 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1117 msg_set_nameinst(hdr, 0);
1119 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1120 msg_set_destport(hdr, 0);
1121 msg_set_destnode(hdr, 0);
1122 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1124 /* Avoid getting stuck with repeated forced replicasts */
1125 msg_set_grp_bc_ack_req(hdr, ack);
1127 /* Build message as chain of buffers */
1128 __skb_queue_head_init(&pkts);
1129 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1130 if (unlikely(rc != dlen))
1134 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1138 /* Update broadcast sequence number and send windows */
1139 tipc_group_update_bc_members(tsk->group, blks, ack);
1141 /* Broadcast link is now free to choose method for next broadcast */
1142 method->mandatory = false;
1143 method->expires = jiffies;
1149 * tipc_send_group_mcast - send message to all members with given identity
1150 * @sock: socket structure
1151 * @m: message to send
1152 * @dlen: total length of message data
1153 * @timeout: timeout to wait for wakeup
1155 * Called from function tipc_sendmsg(), which has done all sanity checks
1156 * Return: the number of bytes sent on success, or errno
1158 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1159 int dlen, long timeout)
1161 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1162 struct sock *sk = sock->sk;
1163 struct tipc_sock *tsk = tipc_sk(sk);
1164 struct tipc_group *grp = tsk->group;
1165 struct tipc_msg *hdr = &tsk->phdr;
1166 struct net *net = sock_net(sk);
1167 struct list_head dsts;
1168 u32 dstcnt, exclude;
1170 INIT_LIST_HEAD(&dsts);
1171 ua->sa.type = msg_nametype(hdr);
1172 ua->scope = msg_lookup_scope(hdr);
1173 exclude = tipc_group_exclude(grp);
1175 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1176 return -EHOSTUNREACH;
1179 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1180 return tipc_send_group_unicast(sock, m, dlen, timeout);
1183 tipc_dest_list_purge(&dsts);
1184 return tipc_send_group_bcast(sock, m, dlen, timeout);
1188 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1189 * @net: the associated network namespace
1190 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1191 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1193 * Multi-threaded: parallel calls with reference to same queues may occur
1195 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1196 struct sk_buff_head *inputq)
1198 u32 self = tipc_own_addr(net);
1199 struct sk_buff *skb, *_skb;
1201 struct sk_buff_head tmpq;
1202 struct list_head dports;
1203 struct tipc_msg *hdr;
1204 struct tipc_uaddr ua;
1205 int user, mtyp, hlen;
1208 __skb_queue_head_init(&tmpq);
1209 INIT_LIST_HEAD(&dports);
1210 ua.addrtype = TIPC_SERVICE_RANGE;
1212 skb = tipc_skb_peek(arrvq, &inputq->lock);
1213 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1215 user = msg_user(hdr);
1216 mtyp = msg_type(hdr);
1217 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1218 onode = msg_orignode(hdr);
1219 ua.sr.type = msg_nametype(hdr);
1221 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1222 spin_lock_bh(&inputq->lock);
1223 if (skb_peek(arrvq) == skb) {
1224 __skb_dequeue(arrvq);
1225 __skb_queue_tail(inputq, skb);
1228 spin_unlock_bh(&inputq->lock);
1232 /* Group messages require exact scope match */
1233 if (msg_in_group(hdr)) {
1236 ua.scope = msg_lookup_scope(hdr);
1239 /* TIPC_NODE_SCOPE means "any scope" in this context */
1241 ua.scope = TIPC_NODE_SCOPE;
1243 ua.scope = TIPC_CLUSTER_SCOPE;
1245 ua.sr.lower = msg_namelower(hdr);
1246 ua.sr.upper = msg_nameupper(hdr);
1249 /* Create destination port list: */
1250 tipc_nametbl_lookup_mcast_sockets(net, &ua, exact, &dports);
1252 /* Clone message per destination */
1253 while (tipc_dest_pop(&dports, NULL, &portid)) {
1254 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1256 msg_set_destport(buf_msg(_skb), portid);
1257 __skb_queue_tail(&tmpq, _skb);
1260 pr_warn("Failed to clone mcast rcv buffer\n");
1262 /* Append to inputq if not already done by other thread */
1263 spin_lock_bh(&inputq->lock);
1264 if (skb_peek(arrvq) == skb) {
1265 skb_queue_splice_tail_init(&tmpq, inputq);
1266 /* Decrease the skb's refcnt as increasing in the
1267 * function tipc_skb_peek
1269 kfree_skb(__skb_dequeue(arrvq));
1271 spin_unlock_bh(&inputq->lock);
1272 __skb_queue_purge(&tmpq);
1275 tipc_sk_rcv(net, inputq);
1278 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1279 * when socket is in Nagle mode
1281 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1283 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1284 struct sk_buff *skb = skb_peek_tail(txq);
1285 struct net *net = sock_net(&tsk->sk);
1286 u32 dnode = tsk_peer_node(tsk);
1290 tsk->pkt_cnt += skb_queue_len(txq);
1291 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1293 if (tsk->nagle_start < NAGLE_START_MAX)
1294 tsk->nagle_start *= 2;
1295 tsk->expect_ack = false;
1296 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1297 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1300 tsk->nagle_start = NAGLE_START_INIT;
1302 msg_set_ack_required(buf_msg(skb));
1303 tsk->expect_ack = true;
1305 tsk->expect_ack = false;
1312 if (!skb || tsk->cong_link_cnt)
1315 /* Do not send SYN again after congestion */
1316 if (msg_is_syn(buf_msg(skb)))
1320 tsk->pkt_cnt += skb_queue_len(txq);
1321 tsk->snt_unacked += tsk->snd_backlog;
1322 tsk->snd_backlog = 0;
1323 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1324 if (rc == -ELINKCONG)
1325 tsk->cong_link_cnt = 1;
1329 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1330 * @tsk: receiving socket
1331 * @skb: pointer to message buffer.
1332 * @inputq: buffer list containing the buffers
1333 * @xmitq: output message area
1335 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1336 struct sk_buff_head *inputq,
1337 struct sk_buff_head *xmitq)
1339 struct tipc_msg *hdr = buf_msg(skb);
1340 u32 onode = tsk_own_node(tsk);
1341 struct sock *sk = &tsk->sk;
1342 int mtyp = msg_type(hdr);
1345 /* Ignore if connection cannot be validated: */
1346 if (!tsk_peer_msg(tsk, hdr)) {
1347 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1351 if (unlikely(msg_errcode(hdr))) {
1352 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1353 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1354 tsk_peer_port(tsk));
1355 sk->sk_state_change(sk);
1357 /* State change is ignored if socket already awake,
1358 * - convert msg to abort msg and add to inqueue
1360 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1361 msg_set_type(hdr, TIPC_CONN_MSG);
1362 msg_set_size(hdr, BASIC_H_SIZE);
1363 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1364 __skb_queue_tail(inputq, skb);
1368 tsk->probe_unacked = false;
1370 if (mtyp == CONN_PROBE) {
1371 msg_set_type(hdr, CONN_PROBE_REPLY);
1372 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1373 __skb_queue_tail(xmitq, skb);
1375 } else if (mtyp == CONN_ACK) {
1376 was_cong = tsk_conn_cong(tsk);
1377 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1378 tsk->snt_unacked -= msg_conn_ack(hdr);
1379 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1380 tsk->snd_win = msg_adv_win(hdr);
1381 if (was_cong && !tsk_conn_cong(tsk))
1382 sk->sk_write_space(sk);
1383 } else if (mtyp != CONN_PROBE_REPLY) {
1384 pr_warn("Received unknown CONN_PROTO msg\n");
1391 * tipc_sendmsg - send message in connectionless manner
1392 * @sock: socket structure
1393 * @m: message to send
1394 * @dsz: amount of user data to be sent
1396 * Message must have an destination specified explicitly.
1397 * Used for SOCK_RDM and SOCK_DGRAM messages,
1398 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1399 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1401 * Return: the number of bytes sent on success, or errno otherwise
1403 static int tipc_sendmsg(struct socket *sock,
1404 struct msghdr *m, size_t dsz)
1406 struct sock *sk = sock->sk;
1410 ret = __tipc_sendmsg(sock, m, dsz);
1416 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1418 struct sock *sk = sock->sk;
1419 struct net *net = sock_net(sk);
1420 struct tipc_sock *tsk = tipc_sk(sk);
1421 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1422 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1423 struct list_head *clinks = &tsk->cong_links;
1424 bool syn = !tipc_sk_type_connectionless(sk);
1425 struct tipc_group *grp = tsk->group;
1426 struct tipc_msg *hdr = &tsk->phdr;
1427 struct tipc_socket_addr skaddr;
1428 struct sk_buff_head pkts;
1431 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1435 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1437 atype = ua->addrtype;
1440 /* If socket belongs to a communication group follow other paths */
1443 return tipc_send_group_bcast(sock, m, dlen, timeout);
1444 if (atype == TIPC_SERVICE_ADDR)
1445 return tipc_send_group_anycast(sock, m, dlen, timeout);
1446 if (atype == TIPC_SOCKET_ADDR)
1447 return tipc_send_group_unicast(sock, m, dlen, timeout);
1448 if (atype == TIPC_SERVICE_RANGE)
1449 return tipc_send_group_mcast(sock, m, dlen, timeout);
1454 ua = (struct tipc_uaddr *)&tsk->peer;
1455 if (!syn && ua->family != AF_TIPC)
1456 return -EDESTADDRREQ;
1457 atype = ua->addrtype;
1460 if (unlikely(syn)) {
1461 if (sk->sk_state == TIPC_LISTEN)
1463 if (sk->sk_state != TIPC_OPEN)
1467 if (atype == TIPC_SERVICE_ADDR) {
1468 tsk->conn_type = ua->sa.type;
1469 tsk->conn_instance = ua->sa.instance;
1471 msg_set_syn(hdr, 1);
1474 /* Determine destination */
1475 if (atype == TIPC_SERVICE_RANGE) {
1476 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1477 } else if (atype == TIPC_SERVICE_ADDR) {
1478 skaddr.node = ua->lookup_node;
1479 ua->scope = tipc_node2scope(skaddr.node);
1480 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1481 return -EHOSTUNREACH;
1482 } else if (atype == TIPC_SOCKET_ADDR) {
1488 /* Block or return if destination link is congested */
1489 rc = tipc_wait_for_cond(sock, &timeout,
1490 !tipc_dest_find(clinks, skaddr.node, 0));
1494 /* Finally build message header */
1495 msg_set_destnode(hdr, skaddr.node);
1496 msg_set_destport(hdr, skaddr.ref);
1497 if (atype == TIPC_SERVICE_ADDR) {
1498 msg_set_type(hdr, TIPC_NAMED_MSG);
1499 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1500 msg_set_nametype(hdr, ua->sa.type);
1501 msg_set_nameinst(hdr, ua->sa.instance);
1502 msg_set_lookup_scope(hdr, ua->scope);
1503 } else { /* TIPC_SOCKET_ADDR */
1504 msg_set_type(hdr, TIPC_DIRECT_MSG);
1505 msg_set_lookup_scope(hdr, 0);
1506 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1509 /* Add message body */
1510 __skb_queue_head_init(&pkts);
1511 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1512 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1513 if (unlikely(rc != dlen))
1515 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1516 __skb_queue_purge(&pkts);
1521 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1522 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1523 if (unlikely(rc == -ELINKCONG)) {
1524 tipc_dest_push(clinks, skaddr.node, 0);
1525 tsk->cong_link_cnt++;
1529 if (unlikely(syn && !rc)) {
1530 tipc_set_sk_state(sk, TIPC_CONNECTING);
1531 if (dlen && timeout) {
1532 timeout = msecs_to_jiffies(timeout);
1533 tipc_wait_for_connect(sock, &timeout);
1537 return rc ? rc : dlen;
1541 * tipc_sendstream - send stream-oriented data
1542 * @sock: socket structure
1544 * @dsz: total length of data to be transmitted
1546 * Used for SOCK_STREAM data.
1548 * Return: the number of bytes sent on success (or partial success),
1549 * or errno if no data sent
1551 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1553 struct sock *sk = sock->sk;
1557 ret = __tipc_sendstream(sock, m, dsz);
1563 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1565 struct sock *sk = sock->sk;
1566 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1567 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1568 struct sk_buff_head *txq = &sk->sk_write_queue;
1569 struct tipc_sock *tsk = tipc_sk(sk);
1570 struct tipc_msg *hdr = &tsk->phdr;
1571 struct net *net = sock_net(sk);
1572 struct sk_buff *skb;
1573 u32 dnode = tsk_peer_node(tsk);
1574 int maxnagle = tsk->maxnagle;
1575 int maxpkt = tsk->max_pkt;
1579 if (unlikely(dlen > INT_MAX))
1582 /* Handle implicit connection setup */
1583 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1584 rc = __tipc_sendmsg(sock, m, dlen);
1585 if (dlen && dlen == rc) {
1586 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1587 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1593 rc = tipc_wait_for_cond(sock, &timeout,
1594 (!tsk->cong_link_cnt &&
1595 !tsk_conn_cong(tsk) &&
1596 tipc_sk_connected(sk)));
1599 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1600 blocks = tsk->snd_backlog;
1601 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1603 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1604 if (unlikely(rc < 0))
1608 if (blocks <= 64 && tsk->expect_ack) {
1609 tsk->snd_backlog = blocks;
1612 } else if (blocks > 64) {
1613 tsk->pkt_cnt += skb_queue_len(txq);
1615 skb = skb_peek_tail(txq);
1617 msg_set_ack_required(buf_msg(skb));
1618 tsk->expect_ack = true;
1620 tsk->expect_ack = false;
1626 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1627 if (unlikely(rc != send))
1629 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1631 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1632 TIPC_DUMP_SK_SNDQ, " ");
1633 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1634 if (unlikely(rc == -ELINKCONG)) {
1635 tsk->cong_link_cnt = 1;
1639 tsk->snt_unacked += blocks;
1640 tsk->snd_backlog = 0;
1643 } while (sent < dlen && !rc);
1645 return sent ? sent : rc;
1649 * tipc_send_packet - send a connection-oriented message
1650 * @sock: socket structure
1651 * @m: message to send
1652 * @dsz: length of data to be transmitted
1654 * Used for SOCK_SEQPACKET messages.
1656 * Return: the number of bytes sent on success, or errno otherwise
1658 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1660 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1663 return tipc_sendstream(sock, m, dsz);
1666 /* tipc_sk_finish_conn - complete the setup of a connection
1668 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1671 struct sock *sk = &tsk->sk;
1672 struct net *net = sock_net(sk);
1673 struct tipc_msg *msg = &tsk->phdr;
1675 msg_set_syn(msg, 0);
1676 msg_set_destnode(msg, peer_node);
1677 msg_set_destport(msg, peer_port);
1678 msg_set_type(msg, TIPC_CONN_MSG);
1679 msg_set_lookup_scope(msg, 0);
1680 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1682 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1683 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1684 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1685 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1686 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1688 __skb_queue_purge(&sk->sk_write_queue);
1689 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1692 /* Fall back to message based flow control */
1693 tsk->rcv_win = FLOWCTL_MSG_WIN;
1694 tsk->snd_win = FLOWCTL_MSG_WIN;
1698 * tipc_sk_set_orig_addr - capture sender's address for received message
1699 * @m: descriptor for message info
1700 * @skb: received message
1702 * Note: Address is not captured if not requested by receiver.
1704 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1706 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1707 struct tipc_msg *hdr = buf_msg(skb);
1712 srcaddr->sock.family = AF_TIPC;
1713 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1714 srcaddr->sock.scope = 0;
1715 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1716 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1717 srcaddr->sock.addr.name.domain = 0;
1718 m->msg_namelen = sizeof(struct sockaddr_tipc);
1720 if (!msg_in_group(hdr))
1723 /* Group message users may also want to know sending member's id */
1724 srcaddr->member.family = AF_TIPC;
1725 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1726 srcaddr->member.scope = 0;
1727 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1728 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1729 srcaddr->member.addr.name.domain = 0;
1730 m->msg_namelen = sizeof(*srcaddr);
1734 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1735 * @m: descriptor for message info
1736 * @skb: received message buffer
1737 * @tsk: TIPC port associated with message
1739 * Note: Ancillary data is not captured if not requested by receiver.
1741 * Return: 0 if successful, otherwise errno
1743 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1744 struct tipc_sock *tsk)
1746 struct tipc_msg *msg;
1753 if (likely(m->msg_controllen == 0))
1757 /* Optionally capture errored message object(s) */
1758 err = msg ? msg_errcode(msg) : 0;
1759 if (unlikely(err)) {
1761 anc_data[1] = msg_data_sz(msg);
1762 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1766 if (skb_linearize(skb))
1769 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1776 /* Optionally capture message destination object */
1777 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1778 switch (dest_type) {
1779 case TIPC_NAMED_MSG:
1781 anc_data[0] = msg_nametype(msg);
1782 anc_data[1] = msg_namelower(msg);
1783 anc_data[2] = msg_namelower(msg);
1785 case TIPC_MCAST_MSG:
1787 anc_data[0] = msg_nametype(msg);
1788 anc_data[1] = msg_namelower(msg);
1789 anc_data[2] = msg_nameupper(msg);
1792 has_name = (tsk->conn_type != 0);
1793 anc_data[0] = tsk->conn_type;
1794 anc_data[1] = tsk->conn_instance;
1795 anc_data[2] = tsk->conn_instance;
1801 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1809 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1811 struct sock *sk = &tsk->sk;
1812 struct sk_buff *skb = NULL;
1813 struct tipc_msg *msg;
1814 u32 peer_port = tsk_peer_port(tsk);
1815 u32 dnode = tsk_peer_node(tsk);
1817 if (!tipc_sk_connected(sk))
1819 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1820 dnode, tsk_own_node(tsk), peer_port,
1821 tsk->portid, TIPC_OK);
1825 msg_set_conn_ack(msg, tsk->rcv_unacked);
1826 tsk->rcv_unacked = 0;
1828 /* Adjust to and advertize the correct window limit */
1829 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1830 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1831 msg_set_adv_win(msg, tsk->rcv_win);
1836 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1838 struct sk_buff *skb;
1840 skb = tipc_sk_build_ack(tsk);
1844 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1845 msg_link_selector(buf_msg(skb)));
1848 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1850 struct sock *sk = sock->sk;
1851 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1852 long timeo = *timeop;
1853 int err = sock_error(sk);
1859 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1860 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1864 add_wait_queue(sk_sleep(sk), &wait);
1866 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1867 sched_annotate_sleep();
1869 remove_wait_queue(sk_sleep(sk), &wait);
1872 if (!skb_queue_empty(&sk->sk_receive_queue))
1877 err = sock_intr_errno(timeo);
1878 if (signal_pending(current))
1881 err = sock_error(sk);
1890 * tipc_recvmsg - receive packet-oriented message
1891 * @sock: network socket
1892 * @m: descriptor for message info
1893 * @buflen: length of user buffer area
1894 * @flags: receive flags
1896 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1897 * If the complete message doesn't fit in user area, truncate it.
1899 * Return: size of returned message data, errno otherwise
1901 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1902 size_t buflen, int flags)
1904 struct sock *sk = sock->sk;
1905 bool connected = !tipc_sk_type_connectionless(sk);
1906 struct tipc_sock *tsk = tipc_sk(sk);
1907 int rc, err, hlen, dlen, copy;
1908 struct sk_buff_head xmitq;
1909 struct tipc_msg *hdr;
1910 struct sk_buff *skb;
1914 /* Catch invalid receive requests */
1915 if (unlikely(!buflen))
1919 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1923 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1925 /* Step rcv queue to first msg with data or error; wait if necessary */
1927 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1930 skb = skb_peek(&sk->sk_receive_queue);
1932 dlen = msg_data_sz(hdr);
1933 hlen = msg_hdr_sz(hdr);
1934 err = msg_errcode(hdr);
1935 grp_evt = msg_is_grp_evt(hdr);
1936 if (likely(dlen || err))
1938 tsk_advance_rx_queue(sk);
1941 /* Collect msg meta data, including error code and rejected data */
1942 tipc_sk_set_orig_addr(m, skb);
1943 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1948 /* Capture data if non-error msg, otherwise just set return value */
1950 copy = min_t(int, dlen, buflen);
1951 if (unlikely(copy != dlen))
1952 m->msg_flags |= MSG_TRUNC;
1953 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1957 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1963 /* Mark message as group event if applicable */
1964 if (unlikely(grp_evt)) {
1965 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1966 m->msg_flags |= MSG_EOR;
1967 m->msg_flags |= MSG_OOB;
1971 /* Caption of data or error code/rejected data was successful */
1972 if (unlikely(flags & MSG_PEEK))
1975 /* Send group flow control advertisement when applicable */
1976 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1977 __skb_queue_head_init(&xmitq);
1978 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1979 msg_orignode(hdr), msg_origport(hdr),
1981 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1984 tsk_advance_rx_queue(sk);
1986 if (likely(!connected))
1989 /* Send connection flow control advertisement when applicable */
1990 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1991 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1992 tipc_sk_send_ack(tsk);
1995 return rc ? rc : copy;
1999 * tipc_recvstream - receive stream-oriented data
2000 * @sock: network socket
2001 * @m: descriptor for message info
2002 * @buflen: total size of user buffer area
2003 * @flags: receive flags
2005 * Used for SOCK_STREAM messages only. If not enough data is available
2006 * will optionally wait for more; never truncates data.
2008 * Return: size of returned message data, errno otherwise
2010 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2011 size_t buflen, int flags)
2013 struct sock *sk = sock->sk;
2014 struct tipc_sock *tsk = tipc_sk(sk);
2015 struct sk_buff *skb;
2016 struct tipc_msg *hdr;
2017 struct tipc_skb_cb *skb_cb;
2018 bool peek = flags & MSG_PEEK;
2019 int offset, required, copy, copied = 0;
2020 int hlen, dlen, err, rc;
2023 /* Catch invalid receive attempts */
2024 if (unlikely(!buflen))
2029 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2033 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2034 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2037 /* Look at first msg in receive queue; wait if necessary */
2038 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2041 skb = skb_peek(&sk->sk_receive_queue);
2042 skb_cb = TIPC_SKB_CB(skb);
2044 dlen = msg_data_sz(hdr);
2045 hlen = msg_hdr_sz(hdr);
2046 err = msg_errcode(hdr);
2048 /* Discard any empty non-errored (SYN-) message */
2049 if (unlikely(!dlen && !err)) {
2050 tsk_advance_rx_queue(sk);
2054 /* Collect msg meta data, incl. error code and rejected data */
2056 tipc_sk_set_orig_addr(m, skb);
2057 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2063 /* Copy data if msg ok, otherwise return error/partial data */
2065 offset = skb_cb->bytes_read;
2066 copy = min_t(int, dlen - offset, buflen - copied);
2067 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2072 if (unlikely(offset < dlen)) {
2074 skb_cb->bytes_read = offset;
2079 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2088 tsk_advance_rx_queue(sk);
2090 /* Send connection flow control advertisement when applicable */
2091 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2092 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2093 tipc_sk_send_ack(tsk);
2095 /* Exit if all requested data or FIN/error received */
2096 if (copied == buflen || err)
2099 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2102 return copied ? copied : rc;
2106 * tipc_write_space - wake up thread if port congestion is released
2109 static void tipc_write_space(struct sock *sk)
2111 struct socket_wq *wq;
2114 wq = rcu_dereference(sk->sk_wq);
2115 if (skwq_has_sleeper(wq))
2116 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2117 EPOLLWRNORM | EPOLLWRBAND);
2122 * tipc_data_ready - wake up threads to indicate messages have been received
2125 static void tipc_data_ready(struct sock *sk)
2127 struct socket_wq *wq;
2130 wq = rcu_dereference(sk->sk_wq);
2131 if (skwq_has_sleeper(wq))
2132 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2133 EPOLLRDNORM | EPOLLRDBAND);
2137 static void tipc_sock_destruct(struct sock *sk)
2139 __skb_queue_purge(&sk->sk_receive_queue);
2142 static void tipc_sk_proto_rcv(struct sock *sk,
2143 struct sk_buff_head *inputq,
2144 struct sk_buff_head *xmitq)
2146 struct sk_buff *skb = __skb_dequeue(inputq);
2147 struct tipc_sock *tsk = tipc_sk(sk);
2148 struct tipc_msg *hdr = buf_msg(skb);
2149 struct tipc_group *grp = tsk->group;
2150 bool wakeup = false;
2152 switch (msg_user(hdr)) {
2154 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2157 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2158 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2160 tsk->cong_link_cnt--;
2162 tipc_sk_push_backlog(tsk, false);
2164 case GROUP_PROTOCOL:
2165 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2168 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2169 hdr, inputq, xmitq);
2176 sk->sk_write_space(sk);
2182 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2184 * @skb: pointer to message buffer.
2185 * @xmitq: for Nagle ACK if any
2186 * Return: true if message should be added to receive queue, false otherwise
2188 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2189 struct sk_buff_head *xmitq)
2191 struct sock *sk = &tsk->sk;
2192 struct net *net = sock_net(sk);
2193 struct tipc_msg *hdr = buf_msg(skb);
2194 bool con_msg = msg_connected(hdr);
2195 u32 pport = tsk_peer_port(tsk);
2196 u32 pnode = tsk_peer_node(tsk);
2197 u32 oport = msg_origport(hdr);
2198 u32 onode = msg_orignode(hdr);
2199 int err = msg_errcode(hdr);
2200 unsigned long delay;
2202 if (unlikely(msg_mcast(hdr)))
2206 switch (sk->sk_state) {
2207 case TIPC_CONNECTING:
2209 if (likely(con_msg)) {
2212 tipc_sk_finish_conn(tsk, oport, onode);
2213 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2214 /* ACK+ message with data is added to receive queue */
2215 if (msg_data_sz(hdr))
2217 /* Empty ACK-, - wake up sleeping connect() and drop */
2218 sk->sk_state_change(sk);
2219 msg_set_dest_droppable(hdr, 1);
2222 /* Ignore connectionless message if not from listening socket */
2223 if (oport != pport || onode != pnode)
2227 if (err != TIPC_ERR_OVERLOAD)
2230 /* Prepare for new setup attempt if we have a SYN clone */
2231 if (skb_queue_empty(&sk->sk_write_queue))
2233 get_random_bytes(&delay, 2);
2234 delay %= (tsk->conn_timeout / 4);
2235 delay = msecs_to_jiffies(delay + 100);
2236 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2239 case TIPC_DISCONNECTING:
2242 /* Accept only SYN message */
2243 if (!msg_is_syn(hdr) &&
2244 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2246 if (!con_msg && !err)
2249 case TIPC_ESTABLISHED:
2250 if (!skb_queue_empty(&sk->sk_write_queue))
2251 tipc_sk_push_backlog(tsk, false);
2252 /* Accept only connection-based messages sent by peer */
2253 if (likely(con_msg && !err && pport == oport &&
2255 if (msg_ack_required(hdr)) {
2256 struct sk_buff *skb;
2258 skb = tipc_sk_build_ack(tsk);
2260 msg_set_nagle_ack(buf_msg(skb));
2261 __skb_queue_tail(xmitq, skb);
2266 if (!tsk_peer_msg(tsk, hdr))
2270 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2271 tipc_node_remove_conn(net, pnode, tsk->portid);
2272 sk->sk_state_change(sk);
2275 pr_err("Unknown sk_state %u\n", sk->sk_state);
2277 /* Abort connection setup attempt */
2278 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2279 sk->sk_err = ECONNREFUSED;
2280 sk->sk_state_change(sk);
2285 * rcvbuf_limit - get proper overload limit of socket receive queue
2289 * For connection oriented messages, irrespective of importance,
2290 * default queue limit is 2 MB.
2292 * For connectionless messages, queue limits are based on message
2293 * importance as follows:
2295 * TIPC_LOW_IMPORTANCE (2 MB)
2296 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2297 * TIPC_HIGH_IMPORTANCE (8 MB)
2298 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2300 * Return: overload limit according to corresponding message importance
2302 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2304 struct tipc_sock *tsk = tipc_sk(sk);
2305 struct tipc_msg *hdr = buf_msg(skb);
2307 if (unlikely(msg_in_group(hdr)))
2308 return READ_ONCE(sk->sk_rcvbuf);
2310 if (unlikely(!msg_connected(hdr)))
2311 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2313 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2314 return READ_ONCE(sk->sk_rcvbuf);
2316 return FLOWCTL_MSG_LIM;
2320 * tipc_sk_filter_rcv - validate incoming message
2322 * @skb: pointer to message.
2323 * @xmitq: output message area (FIXME)
2325 * Enqueues message on receive queue if acceptable; optionally handles
2326 * disconnect indication for a connected socket.
2328 * Called with socket lock already taken
2330 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2331 struct sk_buff_head *xmitq)
2333 bool sk_conn = !tipc_sk_type_connectionless(sk);
2334 struct tipc_sock *tsk = tipc_sk(sk);
2335 struct tipc_group *grp = tsk->group;
2336 struct tipc_msg *hdr = buf_msg(skb);
2337 struct net *net = sock_net(sk);
2338 struct sk_buff_head inputq;
2339 int mtyp = msg_type(hdr);
2340 int limit, err = TIPC_OK;
2342 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2343 TIPC_SKB_CB(skb)->bytes_read = 0;
2344 __skb_queue_head_init(&inputq);
2345 __skb_queue_tail(&inputq, skb);
2347 if (unlikely(!msg_isdata(hdr)))
2348 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2351 tipc_group_filter_msg(grp, &inputq, xmitq);
2353 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2354 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2356 /* Validate and add to receive buffer if there is space */
2357 while ((skb = __skb_dequeue(&inputq))) {
2359 limit = rcvbuf_limit(sk, skb);
2360 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2361 (!sk_conn && msg_connected(hdr)) ||
2362 (!grp && msg_in_group(hdr)))
2363 err = TIPC_ERR_NO_PORT;
2364 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2365 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2367 atomic_inc(&sk->sk_drops);
2368 err = TIPC_ERR_OVERLOAD;
2371 if (unlikely(err)) {
2372 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2373 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2375 __skb_queue_tail(xmitq, skb);
2380 __skb_queue_tail(&sk->sk_receive_queue, skb);
2381 skb_set_owner_r(skb, sk);
2382 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2383 "rcvq >90% allocated!");
2384 sk->sk_data_ready(sk);
2389 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2393 * Caller must hold socket lock
2395 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2397 unsigned int before = sk_rmem_alloc_get(sk);
2398 struct sk_buff_head xmitq;
2401 __skb_queue_head_init(&xmitq);
2403 tipc_sk_filter_rcv(sk, skb, &xmitq);
2404 added = sk_rmem_alloc_get(sk) - before;
2405 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2407 /* Send pending response/rejected messages, if any */
2408 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2413 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2414 * inputq and try adding them to socket or backlog queue
2415 * @inputq: list of incoming buffers with potentially different destinations
2416 * @sk: socket where the buffers should be enqueued
2417 * @dport: port number for the socket
2418 * @xmitq: output queue
2420 * Caller must hold socket lock
2422 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2423 u32 dport, struct sk_buff_head *xmitq)
2425 unsigned long time_limit = jiffies + 2;
2426 struct sk_buff *skb;
2431 while (skb_queue_len(inputq)) {
2432 if (unlikely(time_after_eq(jiffies, time_limit)))
2435 skb = tipc_skb_dequeue(inputq, dport);
2439 /* Add message directly to receive queue if possible */
2440 if (!sock_owned_by_user(sk)) {
2441 tipc_sk_filter_rcv(sk, skb, xmitq);
2445 /* Try backlog, compensating for double-counted bytes */
2446 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2447 if (!sk->sk_backlog.len)
2448 atomic_set(dcnt, 0);
2449 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2450 if (likely(!sk_add_backlog(sk, skb, lim))) {
2451 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2452 "bklg & rcvq >90% allocated!");
2456 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2457 /* Overload => reject message back to sender */
2458 onode = tipc_own_addr(sock_net(sk));
2459 atomic_inc(&sk->sk_drops);
2460 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2461 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2463 __skb_queue_tail(xmitq, skb);
2470 * tipc_sk_rcv - handle a chain of incoming buffers
2471 * @net: the associated network namespace
2472 * @inputq: buffer list containing the buffers
2473 * Consumes all buffers in list until inputq is empty
2474 * Note: may be called in multiple threads referring to the same queue
2476 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2478 struct sk_buff_head xmitq;
2479 u32 dnode, dport = 0;
2481 struct tipc_sock *tsk;
2483 struct sk_buff *skb;
2485 __skb_queue_head_init(&xmitq);
2486 while (skb_queue_len(inputq)) {
2487 dport = tipc_skb_peek_port(inputq, dport);
2488 tsk = tipc_sk_lookup(net, dport);
2492 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2493 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2494 spin_unlock_bh(&sk->sk_lock.slock);
2496 /* Send pending response/rejected messages, if any */
2497 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2501 /* No destination socket => dequeue skb if still there */
2502 skb = tipc_skb_dequeue(inputq, dport);
2506 /* Try secondary lookup if unresolved named message */
2507 err = TIPC_ERR_NO_PORT;
2508 if (tipc_msg_lookup_dest(net, skb, &err))
2511 /* Prepare for message rejection */
2512 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2515 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2517 dnode = msg_destnode(buf_msg(skb));
2518 tipc_node_xmit_skb(net, skb, dnode, dport);
2522 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2524 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2525 struct sock *sk = sock->sk;
2529 int err = sock_error(sk);
2534 if (signal_pending(current))
2535 return sock_intr_errno(*timeo_p);
2536 if (sk->sk_state == TIPC_DISCONNECTING)
2539 add_wait_queue(sk_sleep(sk), &wait);
2540 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2542 remove_wait_queue(sk_sleep(sk), &wait);
2547 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2549 if (addr->family != AF_TIPC)
2551 if (addr->addrtype == TIPC_SERVICE_RANGE)
2552 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2553 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2554 addr->addrtype == TIPC_SOCKET_ADDR);
2558 * tipc_connect - establish a connection to another TIPC port
2559 * @sock: socket structure
2560 * @dest: socket address for destination port
2561 * @destlen: size of socket address data structure
2562 * @flags: file-related flags associated with socket
2564 * Return: 0 on success, errno otherwise
2566 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2567 int destlen, int flags)
2569 struct sock *sk = sock->sk;
2570 struct tipc_sock *tsk = tipc_sk(sk);
2571 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2572 struct msghdr m = {NULL,};
2573 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2577 if (destlen != sizeof(struct sockaddr_tipc))
2587 if (dst->family == AF_UNSPEC) {
2588 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2589 if (!tipc_sk_type_connectionless(sk))
2593 if (!tipc_sockaddr_is_sane(dst)) {
2597 /* DGRAM/RDM connect(), just save the destaddr */
2598 if (tipc_sk_type_connectionless(sk)) {
2599 memcpy(&tsk->peer, dest, destlen);
2601 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2606 previous = sk->sk_state;
2608 switch (sk->sk_state) {
2610 /* Send a 'SYN-' to destination */
2612 m.msg_namelen = destlen;
2614 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2615 * indicate send_msg() is never blocked.
2618 m.msg_flags = MSG_DONTWAIT;
2620 res = __tipc_sendmsg(sock, &m, 0);
2621 if ((res < 0) && (res != -EWOULDBLOCK))
2624 /* Just entered TIPC_CONNECTING state; the only
2625 * difference is that return value in non-blocking
2626 * case is EINPROGRESS, rather than EALREADY.
2630 case TIPC_CONNECTING:
2632 if (previous == TIPC_CONNECTING)
2636 timeout = msecs_to_jiffies(timeout);
2637 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2638 res = tipc_wait_for_connect(sock, &timeout);
2640 case TIPC_ESTABLISHED:
2653 * tipc_listen - allow socket to listen for incoming connections
2654 * @sock: socket structure
2657 * Return: 0 on success, errno otherwise
2659 static int tipc_listen(struct socket *sock, int len)
2661 struct sock *sk = sock->sk;
2665 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2671 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2673 struct sock *sk = sock->sk;
2674 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2677 /* True wake-one mechanism for incoming connections: only
2678 * one process gets woken up, not the 'whole herd'.
2679 * Since we do not 'race & poll' for established sockets
2680 * anymore, the common case will execute the loop only once.
2683 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2684 add_wait_queue(sk_sleep(sk), &wait);
2686 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2688 remove_wait_queue(sk_sleep(sk), &wait);
2691 if (!skb_queue_empty(&sk->sk_receive_queue))
2696 err = sock_intr_errno(timeo);
2697 if (signal_pending(current))
2704 * tipc_accept - wait for connection request
2705 * @sock: listening socket
2706 * @new_sock: new socket that is to be connected
2707 * @flags: file-related flags associated with socket
2708 * @kern: caused by kernel or by userspace?
2710 * Return: 0 on success, errno otherwise
2712 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2715 struct sock *new_sk, *sk = sock->sk;
2716 struct tipc_sock *new_tsock;
2717 struct msghdr m = {NULL,};
2718 struct tipc_msg *msg;
2719 struct sk_buff *buf;
2725 if (sk->sk_state != TIPC_LISTEN) {
2729 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2730 res = tipc_wait_for_accept(sock, timeo);
2734 buf = skb_peek(&sk->sk_receive_queue);
2736 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2739 security_sk_clone(sock->sk, new_sock->sk);
2741 new_sk = new_sock->sk;
2742 new_tsock = tipc_sk(new_sk);
2745 /* we lock on new_sk; but lockdep sees the lock on sk */
2746 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2749 * Reject any stray messages received by new socket
2750 * before the socket lock was taken (very, very unlikely)
2752 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2754 /* Connect new socket to it's peer */
2755 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2757 tsk_set_importance(new_sk, msg_importance(msg));
2758 if (msg_named(msg)) {
2759 new_tsock->conn_type = msg_nametype(msg);
2760 new_tsock->conn_instance = msg_nameinst(msg);
2764 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2765 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2767 if (!msg_data_sz(msg)) {
2768 tsk_advance_rx_queue(sk);
2770 __skb_dequeue(&sk->sk_receive_queue);
2771 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2772 skb_set_owner_r(buf, new_sk);
2774 __tipc_sendstream(new_sock, &m, 0);
2775 release_sock(new_sk);
2782 * tipc_shutdown - shutdown socket connection
2783 * @sock: socket structure
2784 * @how: direction to close (must be SHUT_RDWR)
2786 * Terminates connection (if necessary), then purges socket's receive queue.
2788 * Return: 0 on success, errno otherwise
2790 static int tipc_shutdown(struct socket *sock, int how)
2792 struct sock *sk = sock->sk;
2795 if (how != SHUT_RDWR)
2800 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2801 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2802 sk->sk_shutdown = SHUTDOWN_MASK;
2804 if (sk->sk_state == TIPC_DISCONNECTING) {
2805 /* Discard any unreceived messages */
2806 __skb_queue_purge(&sk->sk_receive_queue);
2812 /* Wake up anyone sleeping in poll. */
2813 sk->sk_state_change(sk);
2819 static void tipc_sk_check_probing_state(struct sock *sk,
2820 struct sk_buff_head *list)
2822 struct tipc_sock *tsk = tipc_sk(sk);
2823 u32 pnode = tsk_peer_node(tsk);
2824 u32 pport = tsk_peer_port(tsk);
2825 u32 self = tsk_own_node(tsk);
2826 u32 oport = tsk->portid;
2827 struct sk_buff *skb;
2829 if (tsk->probe_unacked) {
2830 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2831 sk->sk_err = ECONNABORTED;
2832 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2833 sk->sk_state_change(sk);
2836 /* Prepare new probe */
2837 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2838 pnode, self, pport, oport, TIPC_OK);
2840 __skb_queue_tail(list, skb);
2841 tsk->probe_unacked = true;
2842 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2845 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2847 struct tipc_sock *tsk = tipc_sk(sk);
2849 /* Try again later if dest link is congested */
2850 if (tsk->cong_link_cnt) {
2851 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2854 /* Prepare SYN for retransmit */
2855 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2858 static void tipc_sk_timeout(struct timer_list *t)
2860 struct sock *sk = from_timer(sk, t, sk_timer);
2861 struct tipc_sock *tsk = tipc_sk(sk);
2862 u32 pnode = tsk_peer_node(tsk);
2863 struct sk_buff_head list;
2866 __skb_queue_head_init(&list);
2869 /* Try again later if socket is busy */
2870 if (sock_owned_by_user(sk)) {
2871 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2877 if (sk->sk_state == TIPC_ESTABLISHED)
2878 tipc_sk_check_probing_state(sk, &list);
2879 else if (sk->sk_state == TIPC_CONNECTING)
2880 tipc_sk_retry_connect(sk, &list);
2884 if (!skb_queue_empty(&list))
2885 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2887 /* SYN messages may cause link congestion */
2888 if (rc == -ELINKCONG) {
2889 tipc_dest_push(&tsk->cong_links, pnode, 0);
2890 tsk->cong_link_cnt = 1;
2895 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2897 struct sock *sk = &tsk->sk;
2898 struct net *net = sock_net(sk);
2899 struct tipc_socket_addr skaddr;
2900 struct publication *p;
2903 if (tipc_sk_connected(sk))
2905 key = tsk->portid + tsk->pub_count + 1;
2906 if (key == tsk->portid)
2908 skaddr.ref = tsk->portid;
2909 skaddr.node = tipc_own_addr(net);
2910 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2914 list_add(&p->binding_sock, &tsk->publications);
2916 tsk->published = true;
2920 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2922 struct net *net = sock_net(&tsk->sk);
2923 struct publication *safe, *p;
2924 struct tipc_uaddr _ua;
2927 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2929 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2930 p->sr.type, p->sr.lower, p->sr.upper);
2931 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2934 /* Unbind specific publication */
2935 if (p->scope != ua->scope)
2937 if (p->sr.type != ua->sr.type)
2939 if (p->sr.lower != ua->sr.lower)
2941 if (p->sr.upper != ua->sr.upper)
2943 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2947 if (list_empty(&tsk->publications)) {
2954 /* tipc_sk_reinit: set non-zero address in all existing sockets
2955 * when we go from standalone to network mode.
2957 void tipc_sk_reinit(struct net *net)
2959 struct tipc_net *tn = net_generic(net, tipc_net_id);
2960 struct rhashtable_iter iter;
2961 struct tipc_sock *tsk;
2962 struct tipc_msg *msg;
2964 rhashtable_walk_enter(&tn->sk_rht, &iter);
2967 rhashtable_walk_start(&iter);
2969 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2970 sock_hold(&tsk->sk);
2971 rhashtable_walk_stop(&iter);
2972 lock_sock(&tsk->sk);
2974 msg_set_prevnode(msg, tipc_own_addr(net));
2975 msg_set_orignode(msg, tipc_own_addr(net));
2976 release_sock(&tsk->sk);
2977 rhashtable_walk_start(&iter);
2981 rhashtable_walk_stop(&iter);
2982 } while (tsk == ERR_PTR(-EAGAIN));
2984 rhashtable_walk_exit(&iter);
2987 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2989 struct tipc_net *tn = net_generic(net, tipc_net_id);
2990 struct tipc_sock *tsk;
2993 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2995 sock_hold(&tsk->sk);
3001 static int tipc_sk_insert(struct tipc_sock *tsk)
3003 struct sock *sk = &tsk->sk;
3004 struct net *net = sock_net(sk);
3005 struct tipc_net *tn = net_generic(net, tipc_net_id);
3006 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3007 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3009 while (remaining--) {
3011 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3012 portid = TIPC_MIN_PORT;
3013 tsk->portid = portid;
3014 sock_hold(&tsk->sk);
3015 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3024 static void tipc_sk_remove(struct tipc_sock *tsk)
3026 struct sock *sk = &tsk->sk;
3027 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3029 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3030 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3035 static const struct rhashtable_params tsk_rht_params = {
3037 .head_offset = offsetof(struct tipc_sock, node),
3038 .key_offset = offsetof(struct tipc_sock, portid),
3039 .key_len = sizeof(u32), /* portid */
3040 .max_size = 1048576,
3042 .automatic_shrinking = true,
3045 int tipc_sk_rht_init(struct net *net)
3047 struct tipc_net *tn = net_generic(net, tipc_net_id);
3049 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3052 void tipc_sk_rht_destroy(struct net *net)
3054 struct tipc_net *tn = net_generic(net, tipc_net_id);
3056 /* Wait for socket readers to complete */
3059 rhashtable_destroy(&tn->sk_rht);
3062 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3064 struct net *net = sock_net(&tsk->sk);
3065 struct tipc_group *grp = tsk->group;
3066 struct tipc_msg *hdr = &tsk->phdr;
3067 struct tipc_uaddr ua;
3070 if (mreq->type < TIPC_RESERVED_TYPES)
3072 if (mreq->scope > TIPC_NODE_SCOPE)
3074 if (mreq->scope != TIPC_NODE_SCOPE)
3075 mreq->scope = TIPC_CLUSTER_SCOPE;
3078 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3082 msg_set_lookup_scope(hdr, mreq->scope);
3083 msg_set_nametype(hdr, mreq->type);
3084 msg_set_dest_droppable(hdr, true);
3085 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3086 mreq->type, mreq->instance, mreq->instance);
3087 tipc_nametbl_build_group(net, grp, &ua);
3088 rc = tipc_sk_publish(tsk, &ua);
3090 tipc_group_delete(net, grp);
3094 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3095 tsk->mc_method.rcast = true;
3096 tsk->mc_method.mandatory = true;
3097 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3101 static int tipc_sk_leave(struct tipc_sock *tsk)
3103 struct net *net = sock_net(&tsk->sk);
3104 struct tipc_group *grp = tsk->group;
3105 struct tipc_uaddr ua;
3110 ua.addrtype = TIPC_SERVICE_RANGE;
3111 tipc_group_self(grp, &ua.sr, &scope);
3113 tipc_group_delete(net, grp);
3115 tipc_sk_withdraw(tsk, &ua);
3120 * tipc_setsockopt - set socket option
3121 * @sock: socket structure
3122 * @lvl: option level
3123 * @opt: option identifier
3124 * @ov: pointer to new option value
3125 * @ol: length of option value
3127 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3128 * (to ease compatibility).
3130 * Return: 0 on success, errno otherwise
3132 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3133 sockptr_t ov, unsigned int ol)
3135 struct sock *sk = sock->sk;
3136 struct tipc_sock *tsk = tipc_sk(sk);
3137 struct tipc_group_req mreq;
3141 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3143 if (lvl != SOL_TIPC)
3144 return -ENOPROTOOPT;
3147 case TIPC_IMPORTANCE:
3148 case TIPC_SRC_DROPPABLE:
3149 case TIPC_DEST_DROPPABLE:
3150 case TIPC_CONN_TIMEOUT:
3152 if (ol < sizeof(value))
3154 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3157 case TIPC_GROUP_JOIN:
3158 if (ol < sizeof(mreq))
3160 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3164 if (!sockptr_is_null(ov) || ol)
3171 case TIPC_IMPORTANCE:
3172 res = tsk_set_importance(sk, value);
3174 case TIPC_SRC_DROPPABLE:
3175 if (sock->type != SOCK_STREAM)
3176 tsk_set_unreliable(tsk, value);
3180 case TIPC_DEST_DROPPABLE:
3181 tsk_set_unreturnable(tsk, value);
3183 case TIPC_CONN_TIMEOUT:
3184 tipc_sk(sk)->conn_timeout = value;
3186 case TIPC_MCAST_BROADCAST:
3187 tsk->mc_method.rcast = false;
3188 tsk->mc_method.mandatory = true;
3190 case TIPC_MCAST_REPLICAST:
3191 tsk->mc_method.rcast = true;
3192 tsk->mc_method.mandatory = true;
3194 case TIPC_GROUP_JOIN:
3195 res = tipc_sk_join(tsk, &mreq);
3197 case TIPC_GROUP_LEAVE:
3198 res = tipc_sk_leave(tsk);
3201 tsk->nodelay = !!value;
3214 * tipc_getsockopt - get socket option
3215 * @sock: socket structure
3216 * @lvl: option level
3217 * @opt: option identifier
3218 * @ov: receptacle for option value
3219 * @ol: receptacle for length of option value
3221 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3222 * (to ease compatibility).
3224 * Return: 0 on success, errno otherwise
3226 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3227 char __user *ov, int __user *ol)
3229 struct sock *sk = sock->sk;
3230 struct tipc_sock *tsk = tipc_sk(sk);
3231 struct tipc_service_range seq;
3236 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3237 return put_user(0, ol);
3238 if (lvl != SOL_TIPC)
3239 return -ENOPROTOOPT;
3240 res = get_user(len, ol);
3247 case TIPC_IMPORTANCE:
3248 value = tsk_importance(tsk);
3250 case TIPC_SRC_DROPPABLE:
3251 value = tsk_unreliable(tsk);
3253 case TIPC_DEST_DROPPABLE:
3254 value = tsk_unreturnable(tsk);
3256 case TIPC_CONN_TIMEOUT:
3257 value = tsk->conn_timeout;
3258 /* no need to set "res", since already 0 at this point */
3260 case TIPC_NODE_RECVQ_DEPTH:
3261 value = 0; /* was tipc_queue_size, now obsolete */
3263 case TIPC_SOCK_RECVQ_DEPTH:
3264 value = skb_queue_len(&sk->sk_receive_queue);
3266 case TIPC_SOCK_RECVQ_USED:
3267 value = sk_rmem_alloc_get(sk);
3269 case TIPC_GROUP_JOIN:
3272 tipc_group_self(tsk->group, &seq, &scope);
3282 return res; /* "get" failed */
3284 if (len < sizeof(value))
3287 if (copy_to_user(ov, &value, sizeof(value)))
3290 return put_user(sizeof(value), ol);
3293 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3295 struct net *net = sock_net(sock->sk);
3296 struct tipc_sioc_nodeid_req nr = {0};
3297 struct tipc_sioc_ln_req lnr;
3298 void __user *argp = (void __user *)arg;
3301 case SIOCGETLINKNAME:
3302 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3304 if (!tipc_node_get_linkname(net,
3305 lnr.bearer_id & 0xffff, lnr.peer,
3306 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3307 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3311 return -EADDRNOTAVAIL;
3313 if (copy_from_user(&nr, argp, sizeof(nr)))
3315 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3316 return -EADDRNOTAVAIL;
3317 if (copy_to_user(argp, &nr, sizeof(nr)))
3321 return -ENOIOCTLCMD;
3325 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3327 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3328 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3329 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3331 tsk1->peer.family = AF_TIPC;
3332 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3333 tsk1->peer.scope = TIPC_NODE_SCOPE;
3334 tsk1->peer.addr.id.ref = tsk2->portid;
3335 tsk1->peer.addr.id.node = onode;
3336 tsk2->peer.family = AF_TIPC;
3337 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3338 tsk2->peer.scope = TIPC_NODE_SCOPE;
3339 tsk2->peer.addr.id.ref = tsk1->portid;
3340 tsk2->peer.addr.id.node = onode;
3342 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3343 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3347 /* Protocol switches for the various types of TIPC sockets */
3349 static const struct proto_ops msg_ops = {
3350 .owner = THIS_MODULE,
3352 .release = tipc_release,
3354 .connect = tipc_connect,
3355 .socketpair = tipc_socketpair,
3356 .accept = sock_no_accept,
3357 .getname = tipc_getname,
3359 .ioctl = tipc_ioctl,
3360 .listen = sock_no_listen,
3361 .shutdown = tipc_shutdown,
3362 .setsockopt = tipc_setsockopt,
3363 .getsockopt = tipc_getsockopt,
3364 .sendmsg = tipc_sendmsg,
3365 .recvmsg = tipc_recvmsg,
3366 .mmap = sock_no_mmap,
3367 .sendpage = sock_no_sendpage
3370 static const struct proto_ops packet_ops = {
3371 .owner = THIS_MODULE,
3373 .release = tipc_release,
3375 .connect = tipc_connect,
3376 .socketpair = tipc_socketpair,
3377 .accept = tipc_accept,
3378 .getname = tipc_getname,
3380 .ioctl = tipc_ioctl,
3381 .listen = tipc_listen,
3382 .shutdown = tipc_shutdown,
3383 .setsockopt = tipc_setsockopt,
3384 .getsockopt = tipc_getsockopt,
3385 .sendmsg = tipc_send_packet,
3386 .recvmsg = tipc_recvmsg,
3387 .mmap = sock_no_mmap,
3388 .sendpage = sock_no_sendpage
3391 static const struct proto_ops stream_ops = {
3392 .owner = THIS_MODULE,
3394 .release = tipc_release,
3396 .connect = tipc_connect,
3397 .socketpair = tipc_socketpair,
3398 .accept = tipc_accept,
3399 .getname = tipc_getname,
3401 .ioctl = tipc_ioctl,
3402 .listen = tipc_listen,
3403 .shutdown = tipc_shutdown,
3404 .setsockopt = tipc_setsockopt,
3405 .getsockopt = tipc_getsockopt,
3406 .sendmsg = tipc_sendstream,
3407 .recvmsg = tipc_recvstream,
3408 .mmap = sock_no_mmap,
3409 .sendpage = sock_no_sendpage
3412 static const struct net_proto_family tipc_family_ops = {
3413 .owner = THIS_MODULE,
3415 .create = tipc_sk_create
3418 static struct proto tipc_proto = {
3420 .owner = THIS_MODULE,
3421 .obj_size = sizeof(struct tipc_sock),
3422 .sysctl_rmem = sysctl_tipc_rmem
3426 * tipc_socket_init - initialize TIPC socket interface
3428 * Return: 0 on success, errno otherwise
3430 int tipc_socket_init(void)
3434 res = proto_register(&tipc_proto, 1);
3436 pr_err("Failed to register TIPC protocol type\n");
3440 res = sock_register(&tipc_family_ops);
3442 pr_err("Failed to register TIPC socket type\n");
3443 proto_unregister(&tipc_proto);
3451 * tipc_socket_stop - stop TIPC socket interface
3453 void tipc_socket_stop(void)
3455 sock_unregister(tipc_family_ops.family);
3456 proto_unregister(&tipc_proto);
3459 /* Caller should hold socket lock for the passed tipc socket. */
3460 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3464 struct nlattr *nest;
3466 peer_node = tsk_peer_node(tsk);
3467 peer_port = tsk_peer_port(tsk);
3469 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3475 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3478 if (tsk->conn_type != 0) {
3479 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3481 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3483 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3486 nla_nest_end(skb, nest);
3491 nla_nest_cancel(skb, nest);
3496 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3499 struct net *net = sock_net(skb->sk);
3500 struct sock *sk = &tsk->sk;
3502 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3503 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3506 if (tipc_sk_connected(sk)) {
3507 if (__tipc_nl_add_sk_con(skb, tsk))
3509 } else if (!list_empty(&tsk->publications)) {
3510 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3516 /* Caller should hold socket lock for the passed tipc socket. */
3517 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3518 struct tipc_sock *tsk)
3520 struct nlattr *attrs;
3523 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3524 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3530 goto genlmsg_cancel;
3532 if (__tipc_nl_add_sk_info(skb, tsk))
3533 goto attr_msg_cancel;
3535 nla_nest_end(skb, attrs);
3536 genlmsg_end(skb, hdr);
3541 nla_nest_cancel(skb, attrs);
3543 genlmsg_cancel(skb, hdr);
3548 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3549 int (*skb_handler)(struct sk_buff *skb,
3550 struct netlink_callback *cb,
3551 struct tipc_sock *tsk))
3553 struct rhashtable_iter *iter = (void *)cb->args[4];
3554 struct tipc_sock *tsk;
3557 rhashtable_walk_start(iter);
3558 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3561 if (err == -EAGAIN) {
3568 sock_hold(&tsk->sk);
3569 rhashtable_walk_stop(iter);
3570 lock_sock(&tsk->sk);
3571 err = skb_handler(skb, cb, tsk);
3573 release_sock(&tsk->sk);
3577 release_sock(&tsk->sk);
3578 rhashtable_walk_start(iter);
3581 rhashtable_walk_stop(iter);
3585 EXPORT_SYMBOL(tipc_nl_sk_walk);
3587 int tipc_dump_start(struct netlink_callback *cb)
3589 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3591 EXPORT_SYMBOL(tipc_dump_start);
3593 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3595 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3596 struct rhashtable_iter *iter = (void *)cb->args[4];
3597 struct tipc_net *tn = tipc_net(net);
3600 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3604 cb->args[4] = (long)iter;
3607 rhashtable_walk_enter(&tn->sk_rht, iter);
3611 int tipc_dump_done(struct netlink_callback *cb)
3613 struct rhashtable_iter *hti = (void *)cb->args[4];
3615 rhashtable_walk_exit(hti);
3619 EXPORT_SYMBOL(tipc_dump_done);
3621 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3622 struct tipc_sock *tsk, u32 sk_filter_state,
3623 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3625 struct sock *sk = &tsk->sk;
3626 struct nlattr *attrs;
3627 struct nlattr *stat;
3629 /*filter response w.r.t sk_state*/
3630 if (!(sk_filter_state & (1 << sk->sk_state)))
3633 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3637 if (__tipc_nl_add_sk_info(skb, tsk))
3638 goto attr_msg_cancel;
3640 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3641 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3642 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3643 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3644 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3646 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3647 tipc_diag_gen_cookie(sk),
3649 goto attr_msg_cancel;
3651 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3653 goto attr_msg_cancel;
3655 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3656 skb_queue_len(&sk->sk_receive_queue)) ||
3657 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3658 skb_queue_len(&sk->sk_write_queue)) ||
3659 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3660 atomic_read(&sk->sk_drops)))
3661 goto stat_msg_cancel;
3663 if (tsk->cong_link_cnt &&
3664 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3665 goto stat_msg_cancel;
3667 if (tsk_conn_cong(tsk) &&
3668 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3669 goto stat_msg_cancel;
3671 nla_nest_end(skb, stat);
3674 if (tipc_group_fill_sock_diag(tsk->group, skb))
3675 goto stat_msg_cancel;
3677 nla_nest_end(skb, attrs);
3682 nla_nest_cancel(skb, stat);
3684 nla_nest_cancel(skb, attrs);
3688 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3690 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3692 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3695 /* Caller should hold socket lock for the passed tipc socket. */
3696 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3697 struct netlink_callback *cb,
3698 struct publication *publ)
3701 struct nlattr *attrs;
3703 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3704 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3708 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3710 goto genlmsg_cancel;
3712 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3713 goto attr_msg_cancel;
3714 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3715 goto attr_msg_cancel;
3716 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3717 goto attr_msg_cancel;
3718 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3719 goto attr_msg_cancel;
3721 nla_nest_end(skb, attrs);
3722 genlmsg_end(skb, hdr);
3727 nla_nest_cancel(skb, attrs);
3729 genlmsg_cancel(skb, hdr);
3734 /* Caller should hold socket lock for the passed tipc socket. */
3735 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3736 struct netlink_callback *cb,
3737 struct tipc_sock *tsk, u32 *last_publ)
3740 struct publication *p;
3743 list_for_each_entry(p, &tsk->publications, binding_sock) {
3744 if (p->key == *last_publ)
3747 if (p->key != *last_publ) {
3748 /* We never set seq or call nl_dump_check_consistent()
3749 * this means that setting prev_seq here will cause the
3750 * consistence check to fail in the netlink callback
3751 * handler. Resulting in the last NLMSG_DONE message
3752 * having the NLM_F_DUMP_INTR flag set.
3759 p = list_first_entry(&tsk->publications, struct publication,
3763 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3764 err = __tipc_nl_add_sk_publ(skb, cb, p);
3766 *last_publ = p->key;
3775 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3778 u32 tsk_portid = cb->args[0];
3779 u32 last_publ = cb->args[1];
3780 u32 done = cb->args[2];
3781 struct net *net = sock_net(skb->sk);
3782 struct tipc_sock *tsk;
3785 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3786 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3788 if (!attrs[TIPC_NLA_SOCK])
3791 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3792 attrs[TIPC_NLA_SOCK],
3793 tipc_nl_sock_policy, NULL);
3797 if (!sock[TIPC_NLA_SOCK_REF])
3800 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3806 tsk = tipc_sk_lookup(net, tsk_portid);
3810 lock_sock(&tsk->sk);
3811 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3814 release_sock(&tsk->sk);
3817 cb->args[0] = tsk_portid;
3818 cb->args[1] = last_publ;
3825 * tipc_sk_filtering - check if a socket should be traced
3826 * @sk: the socket to be examined
3828 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3829 * (portid, sock type, name type, name lower, name upper)
3831 * Return: true if the socket meets the socket tuple data
3832 * (value 0 = 'any') or when there is no tuple set (all = 0),
3835 bool tipc_sk_filtering(struct sock *sk)
3837 struct tipc_sock *tsk;
3838 struct publication *p;
3839 u32 _port, _sktype, _type, _lower, _upper;
3840 u32 type = 0, lower = 0, upper = 0;
3847 _port = sysctl_tipc_sk_filter[0];
3848 _sktype = sysctl_tipc_sk_filter[1];
3849 _type = sysctl_tipc_sk_filter[2];
3850 _lower = sysctl_tipc_sk_filter[3];
3851 _upper = sysctl_tipc_sk_filter[4];
3853 if (!_port && !_sktype && !_type && !_lower && !_upper)
3857 return (_port == tsk->portid);
3859 if (_sktype && _sktype != sk->sk_type)
3862 if (tsk->published) {
3863 p = list_first_entry_or_null(&tsk->publications,
3864 struct publication, binding_sock);
3867 lower = p->sr.lower;
3868 upper = p->sr.upper;
3872 if (!tipc_sk_type_connectionless(sk)) {
3873 type = tsk->conn_type;
3874 lower = tsk->conn_instance;
3875 upper = tsk->conn_instance;
3878 if ((_type && _type != type) || (_lower && _lower != lower) ||
3879 (_upper && _upper != upper))
3885 u32 tipc_sock_get_portid(struct sock *sk)
3887 return (sk) ? (tipc_sk(sk))->portid : 0;
3891 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3892 * both the rcv and backlog queues are considered
3893 * @sk: tipc sk to be checked
3894 * @skb: tipc msg to be checked
3896 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3899 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3901 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3902 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3903 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3905 return (qsize > lim * 90 / 100);
3909 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3910 * only the rcv queue is considered
3911 * @sk: tipc sk to be checked
3912 * @skb: tipc msg to be checked
3914 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3917 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3919 unsigned int lim = rcvbuf_limit(sk, skb);
3920 unsigned int qsize = sk_rmem_alloc_get(sk);
3922 return (qsize > lim * 90 / 100);
3926 * tipc_sk_dump - dump TIPC socket
3927 * @sk: tipc sk to be dumped
3928 * @dqueues: bitmask to decide if any socket queue to be dumped?
3929 * - TIPC_DUMP_NONE: don't dump socket queues
3930 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3931 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3932 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3933 * - TIPC_DUMP_ALL: dump all the socket queues above
3934 * @buf: returned buffer of dump data in format
3936 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3939 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3940 struct tipc_sock *tsk;
3941 struct publication *p;
3945 i += scnprintf(buf, sz, "sk data: (null)\n");
3950 tsk_connected = !tipc_sk_type_connectionless(sk);
3952 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3953 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3954 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3955 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3956 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3957 if (tsk_connected) {
3958 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3959 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3960 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3961 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3963 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3964 if (tsk->published) {
3965 p = list_first_entry_or_null(&tsk->publications,
3966 struct publication, binding_sock);
3967 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3968 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3969 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3971 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3972 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3973 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3974 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3975 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3976 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3977 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3978 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3979 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3980 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3981 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3982 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3983 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3984 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3986 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3987 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3988 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3991 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3992 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3993 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3996 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3997 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3998 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3999 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4000 i += scnprintf(buf + i, sz - i, " tail ");
4001 i += tipc_skb_dump(sk->sk_backlog.tail, false,