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 * @max_pkt: maximum packet size "hint" used when building messages sent by port
77 * @maxnagle: maximum size of msg which can be subject to nagle
78 * @portid: unique port identity in TIPC socket hash table
79 * @phdr: preformatted message header used when sending messages
80 * @cong_links: list of congested links
81 * @publications: list of publications for port
82 * @blocking_link: address of the congested link we are currently sleeping on
83 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @probe_unacked: probe has not received ack yet
86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87 * @cong_link_cnt: number of congested links
88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
89 * @snd_win: send window size
90 * @peer_caps: peer capabilities mask
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @rcv_win: receive window size
93 * @peer: 'connected' peer for dgram/rdm
94 * @node: hash table node
95 * @mc_method: cookie for use between socket and broadcast layer
96 * @rcu: rcu struct for tipc_sock
97 * @group: TIPC communications group
98 * @oneway: message count in one direction (FIXME)
99 * @nagle_start: current nagle value
100 * @snd_backlog: send backlog count
101 * @msg_acc: messages accepted; used in managing backlog and nagle
102 * @pkt_cnt: TIPC socket packet count
103 * @expect_ack: whether this TIPC socket is expecting an ack
104 * @nodelay: setsockopt() TIPC_NODELAY setting
105 * @group_is_open: TIPC socket group is fully open (FIXME)
106 * @published: true if port has one or more associated names
107 * @conn_addrtype: address type used when establishing connection
114 struct tipc_msg phdr;
115 struct list_head cong_links;
116 struct list_head publications;
118 atomic_t dupl_rcvcnt;
127 struct sockaddr_tipc peer;
128 struct rhash_head node;
129 struct tipc_mc_method mc_method;
131 struct tipc_group *group;
144 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145 static void tipc_data_ready(struct sock *sk);
146 static void tipc_write_space(struct sock *sk);
147 static void tipc_sock_destruct(struct sock *sk);
148 static int tipc_release(struct socket *sock);
149 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
151 static void tipc_sk_timeout(struct timer_list *t);
152 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
153 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
154 static int tipc_sk_leave(struct tipc_sock *tsk);
155 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
156 static int tipc_sk_insert(struct tipc_sock *tsk);
157 static void tipc_sk_remove(struct tipc_sock *tsk);
158 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
159 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
160 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
161 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
163 static const struct proto_ops packet_ops;
164 static const struct proto_ops stream_ops;
165 static const struct proto_ops msg_ops;
166 static struct proto tipc_proto;
167 static const struct rhashtable_params tsk_rht_params;
169 static u32 tsk_own_node(struct tipc_sock *tsk)
171 return msg_prevnode(&tsk->phdr);
174 static u32 tsk_peer_node(struct tipc_sock *tsk)
176 return msg_destnode(&tsk->phdr);
179 static u32 tsk_peer_port(struct tipc_sock *tsk)
181 return msg_destport(&tsk->phdr);
184 static bool tsk_unreliable(struct tipc_sock *tsk)
186 return msg_src_droppable(&tsk->phdr) != 0;
189 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
191 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
194 static bool tsk_unreturnable(struct tipc_sock *tsk)
196 return msg_dest_droppable(&tsk->phdr) != 0;
199 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
201 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
204 static int tsk_importance(struct tipc_sock *tsk)
206 return msg_importance(&tsk->phdr);
209 static struct tipc_sock *tipc_sk(const struct sock *sk)
211 return container_of(sk, struct tipc_sock, sk);
214 int tsk_set_importance(struct sock *sk, int imp)
216 if (imp > TIPC_CRITICAL_IMPORTANCE)
218 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
222 static bool tsk_conn_cong(struct tipc_sock *tsk)
224 return tsk->snt_unacked > tsk->snd_win;
227 static u16 tsk_blocks(int len)
229 return ((len / FLOWCTL_BLK_SZ) + 1);
232 /* tsk_blocks(): translate a buffer size in bytes to number of
233 * advertisable blocks, taking into account the ratio truesize(len)/len
234 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
236 static u16 tsk_adv_blocks(int len)
238 return len / FLOWCTL_BLK_SZ / 4;
241 /* tsk_inc(): increment counter for sent or received data
242 * - If block based flow control is not supported by peer we
243 * fall back to message based ditto, incrementing the counter
245 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
247 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
248 return ((msglen / FLOWCTL_BLK_SZ) + 1);
252 /* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
254 static void tsk_set_nagle(struct tipc_sock *tsk)
256 struct sock *sk = &tsk->sk;
259 if (sk->sk_type != SOCK_STREAM)
263 if (!(tsk->peer_caps & TIPC_NAGLE))
265 /* Limit node local buffer size to avoid receive queue overflow */
266 if (tsk->max_pkt == MAX_MSG_SIZE)
267 tsk->maxnagle = 1500;
269 tsk->maxnagle = tsk->max_pkt;
273 * tsk_advance_rx_queue - discard first buffer in socket receive queue
274 * @sk: network socket
276 * Caller must hold socket lock
278 static void tsk_advance_rx_queue(struct sock *sk)
280 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
281 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
284 /* tipc_sk_respond() : send response message back to sender
286 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
290 u32 onode = tipc_own_addr(sock_net(sk));
292 if (!tipc_msg_reverse(onode, &skb, err))
295 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
296 dnode = msg_destnode(buf_msg(skb));
297 selector = msg_origport(buf_msg(skb));
298 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
302 * tsk_rej_rx_queue - reject all buffers in socket receive queue
303 * @sk: network socket
304 * @error: response error code
306 * Caller must hold socket lock
308 static void tsk_rej_rx_queue(struct sock *sk, int error)
312 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
313 tipc_sk_respond(sk, skb, error);
316 static bool tipc_sk_connected(struct sock *sk)
318 return sk->sk_state == TIPC_ESTABLISHED;
321 /* tipc_sk_type_connectionless - check if the socket is datagram socket
324 * Returns true if connection less, false otherwise
326 static bool tipc_sk_type_connectionless(struct sock *sk)
328 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
331 /* tsk_peer_msg - verify if message was sent by connected port's peer
333 * Handles cases where the node's network address has changed from
334 * the default of <0.0.0> to its configured setting.
336 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
338 struct sock *sk = &tsk->sk;
339 u32 self = tipc_own_addr(sock_net(sk));
340 u32 peer_port = tsk_peer_port(tsk);
341 u32 orig_node, peer_node;
343 if (unlikely(!tipc_sk_connected(sk)))
346 if (unlikely(msg_origport(msg) != peer_port))
349 orig_node = msg_orignode(msg);
350 peer_node = tsk_peer_node(tsk);
352 if (likely(orig_node == peer_node))
355 if (!orig_node && peer_node == self)
358 if (!peer_node && orig_node == self)
364 /* tipc_set_sk_state - set the sk_state of the socket
367 * Caller must hold socket lock
369 * Returns 0 on success, errno otherwise
371 static int tipc_set_sk_state(struct sock *sk, int state)
373 int oldsk_state = sk->sk_state;
381 case TIPC_CONNECTING:
382 if (oldsk_state == TIPC_OPEN)
385 case TIPC_ESTABLISHED:
386 if (oldsk_state == TIPC_CONNECTING ||
387 oldsk_state == TIPC_OPEN)
390 case TIPC_DISCONNECTING:
391 if (oldsk_state == TIPC_CONNECTING ||
392 oldsk_state == TIPC_ESTABLISHED)
398 sk->sk_state = state;
403 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
405 struct sock *sk = sock->sk;
406 int err = sock_error(sk);
407 int typ = sock->type;
411 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
412 if (sk->sk_state == TIPC_DISCONNECTING)
414 else if (!tipc_sk_connected(sk))
419 if (signal_pending(current))
420 return sock_intr_errno(*timeout);
425 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
427 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
431 while ((rc_ = !(condition_))) { \
432 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
435 rc_ = tipc_sk_sock_err((sock_), timeo_); \
438 add_wait_queue(sk_sleep(sk_), &wait_); \
440 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
441 sched_annotate_sleep(); \
443 remove_wait_queue(sk_sleep(sk_), &wait_); \
449 * tipc_sk_create - create a TIPC socket
450 * @net: network namespace (must be default network)
451 * @sock: pre-allocated socket structure
452 * @protocol: protocol indicator (must be 0)
453 * @kern: caused by kernel or by userspace?
455 * This routine creates additional data structures used by the TIPC socket,
456 * initializes them, and links them together.
458 * Return: 0 on success, errno otherwise
460 static int tipc_sk_create(struct net *net, struct socket *sock,
461 int protocol, int kern)
463 const struct proto_ops *ops;
465 struct tipc_sock *tsk;
466 struct tipc_msg *msg;
468 /* Validate arguments */
469 if (unlikely(protocol != 0))
470 return -EPROTONOSUPPORT;
472 switch (sock->type) {
487 /* Allocate socket's protocol area */
488 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
493 tsk->max_pkt = MAX_PKT_DEFAULT;
495 tsk->nagle_start = NAGLE_START_INIT;
496 INIT_LIST_HEAD(&tsk->publications);
497 INIT_LIST_HEAD(&tsk->cong_links);
500 /* Finish initializing socket data structures */
502 sock_init_data(sock, sk);
503 tipc_set_sk_state(sk, TIPC_OPEN);
504 if (tipc_sk_insert(tsk)) {
505 pr_warn("Socket create failed; port number exhausted\n");
509 /* Ensure tsk is visible before we read own_addr. */
512 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
513 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
515 msg_set_origport(msg, tsk->portid);
516 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
518 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
519 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
520 sk->sk_data_ready = tipc_data_ready;
521 sk->sk_write_space = tipc_write_space;
522 sk->sk_destruct = tipc_sock_destruct;
523 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
524 tsk->group_is_open = true;
525 atomic_set(&tsk->dupl_rcvcnt, 0);
527 /* Start out with safe limits until we receive an advertised window */
528 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
529 tsk->rcv_win = tsk->snd_win;
531 if (tipc_sk_type_connectionless(sk)) {
532 tsk_set_unreturnable(tsk, true);
533 if (sock->type == SOCK_DGRAM)
534 tsk_set_unreliable(tsk, true);
536 __skb_queue_head_init(&tsk->mc_method.deferredq);
537 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
541 static void tipc_sk_callback(struct rcu_head *head)
543 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
548 /* Caller should hold socket lock for the socket. */
549 static void __tipc_shutdown(struct socket *sock, int error)
551 struct sock *sk = sock->sk;
552 struct tipc_sock *tsk = tipc_sk(sk);
553 struct net *net = sock_net(sk);
554 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
555 u32 dnode = tsk_peer_node(tsk);
558 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
559 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
560 !tsk_conn_cong(tsk)));
562 /* Push out delayed messages if in Nagle mode */
563 tipc_sk_push_backlog(tsk, false);
564 /* Remove pending SYN */
565 __skb_queue_purge(&sk->sk_write_queue);
567 /* Remove partially received buffer if any */
568 skb = skb_peek(&sk->sk_receive_queue);
569 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
570 __skb_unlink(skb, &sk->sk_receive_queue);
574 /* Reject all unreceived messages if connectionless */
575 if (tipc_sk_type_connectionless(sk)) {
576 tsk_rej_rx_queue(sk, error);
580 switch (sk->sk_state) {
581 case TIPC_CONNECTING:
582 case TIPC_ESTABLISHED:
583 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
584 tipc_node_remove_conn(net, dnode, tsk->portid);
585 /* Send a FIN+/- to its peer */
586 skb = __skb_dequeue(&sk->sk_receive_queue);
588 __skb_queue_purge(&sk->sk_receive_queue);
589 tipc_sk_respond(sk, skb, error);
592 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
593 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
594 tsk_own_node(tsk), tsk_peer_port(tsk),
597 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
600 /* Reject all SYN messages */
601 tsk_rej_rx_queue(sk, error);
604 __skb_queue_purge(&sk->sk_receive_queue);
610 * tipc_release - destroy a TIPC socket
611 * @sock: socket to destroy
613 * This routine cleans up any messages that are still queued on the socket.
614 * For DGRAM and RDM socket types, all queued messages are rejected.
615 * For SEQPACKET and STREAM socket types, the first message is rejected
616 * and any others are discarded. (If the first message on a STREAM socket
617 * is partially-read, it is discarded and the next one is rejected instead.)
619 * NOTE: Rejected messages are not necessarily returned to the sender! They
620 * are returned or discarded according to the "destination droppable" setting
621 * specified for the message by the sender.
623 * Return: 0 on success, errno otherwise
625 static int tipc_release(struct socket *sock)
627 struct sock *sk = sock->sk;
628 struct tipc_sock *tsk;
631 * Exit if socket isn't fully initialized (occurs when a failed accept()
632 * releases a pre-allocated child socket that was never used)
640 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
641 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
642 sk->sk_shutdown = SHUTDOWN_MASK;
644 tipc_sk_withdraw(tsk, NULL);
645 __skb_queue_purge(&tsk->mc_method.deferredq);
646 sk_stop_timer(sk, &sk->sk_timer);
650 /* Reject any messages that accumulated in backlog queue */
652 tipc_dest_list_purge(&tsk->cong_links);
653 tsk->cong_link_cnt = 0;
654 call_rcu(&tsk->rcu, tipc_sk_callback);
661 * __tipc_bind - associate or disassocate TIPC name(s) with a socket
662 * @sock: socket structure
663 * @skaddr: socket address describing name(s) and desired operation
664 * @alen: size of socket address data structure
666 * Name and name sequence binding are indicated using a positive scope value;
667 * a negative scope value unbinds the specified name. Specifying no name
668 * (i.e. a socket address length of 0) unbinds all names from the socket.
670 * Return: 0 on success, errno otherwise
672 * NOTE: This routine doesn't need to take the socket lock since it doesn't
673 * access any non-constant socket information.
675 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
677 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
678 struct tipc_sock *tsk = tipc_sk(sock->sk);
682 return tipc_sk_withdraw(tsk, NULL);
684 if (ua->addrtype == TIPC_SERVICE_ADDR) {
685 ua->addrtype = TIPC_SERVICE_RANGE;
686 ua->sr.upper = ua->sr.lower;
690 ua->scope = -ua->scope;
692 /* Users may still use deprecated TIPC_ZONE_SCOPE */
693 if (ua->scope != TIPC_NODE_SCOPE)
694 ua->scope = TIPC_CLUSTER_SCOPE;
700 return tipc_sk_withdraw(tsk, ua);
701 return tipc_sk_publish(tsk, ua);
704 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
709 res = __tipc_bind(sock, skaddr, alen);
710 release_sock(sock->sk);
714 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
716 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
717 u32 atype = ua->addrtype;
720 if (!tipc_uaddr_valid(ua, alen))
722 if (atype == TIPC_SOCKET_ADDR)
723 return -EAFNOSUPPORT;
724 if (ua->sr.type < TIPC_RESERVED_TYPES) {
725 pr_warn_once("Can't bind to reserved service type %u\n",
730 return tipc_sk_bind(sock, skaddr, alen);
734 * tipc_getname - get port ID of socket or peer socket
735 * @sock: socket structure
736 * @uaddr: area for returned socket address
737 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
739 * Return: 0 on success, errno otherwise
741 * NOTE: This routine doesn't need to take the socket lock since it only
742 * accesses socket information that is unchanging (or which changes in
743 * a completely predictable manner).
745 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
748 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
749 struct sock *sk = sock->sk;
750 struct tipc_sock *tsk = tipc_sk(sk);
752 memset(addr, 0, sizeof(*addr));
754 if ((!tipc_sk_connected(sk)) &&
755 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
757 addr->addr.id.ref = tsk_peer_port(tsk);
758 addr->addr.id.node = tsk_peer_node(tsk);
760 addr->addr.id.ref = tsk->portid;
761 addr->addr.id.node = tipc_own_addr(sock_net(sk));
764 addr->addrtype = TIPC_SOCKET_ADDR;
765 addr->family = AF_TIPC;
767 addr->addr.name.domain = 0;
769 return sizeof(*addr);
773 * tipc_poll - read and possibly block on pollmask
774 * @file: file structure associated with the socket
775 * @sock: socket for which to calculate the poll bits
778 * Return: pollmask value
781 * It appears that the usual socket locking mechanisms are not useful here
782 * since the pollmask info is potentially out-of-date the moment this routine
783 * exits. TCP and other protocols seem to rely on higher level poll routines
784 * to handle any preventable race conditions, so TIPC will do the same ...
786 * IMPORTANT: The fact that a read or write operation is indicated does NOT
787 * imply that the operation will succeed, merely that it should be performed
788 * and will not block.
790 static __poll_t tipc_poll(struct file *file, struct socket *sock,
793 struct sock *sk = sock->sk;
794 struct tipc_sock *tsk = tipc_sk(sk);
795 __poll_t revents = 0;
797 sock_poll_wait(file, sock, wait);
798 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
800 if (sk->sk_shutdown & RCV_SHUTDOWN)
801 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
802 if (sk->sk_shutdown == SHUTDOWN_MASK)
805 switch (sk->sk_state) {
806 case TIPC_ESTABLISHED:
807 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
811 case TIPC_CONNECTING:
812 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
813 revents |= EPOLLIN | EPOLLRDNORM;
816 if (tsk->group_is_open && !tsk->cong_link_cnt)
818 if (!tipc_sk_type_connectionless(sk))
820 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
822 revents |= EPOLLIN | EPOLLRDNORM;
824 case TIPC_DISCONNECTING:
825 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
832 * tipc_sendmcast - send multicast message
833 * @sock: socket structure
834 * @ua: destination address struct
835 * @msg: message to send
836 * @dlen: length of data to send
837 * @timeout: timeout to wait for wakeup
839 * Called from function tipc_sendmsg(), which has done all sanity checks
840 * Return: the number of bytes sent on success, or errno
842 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
843 struct msghdr *msg, size_t dlen, long timeout)
845 struct sock *sk = sock->sk;
846 struct tipc_sock *tsk = tipc_sk(sk);
847 struct tipc_msg *hdr = &tsk->phdr;
848 struct net *net = sock_net(sk);
849 int mtu = tipc_bcast_get_mtu(net);
850 struct sk_buff_head pkts;
851 struct tipc_nlist dsts;
857 /* Block or return if any destination link is congested */
858 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
862 /* Lookup destination nodes */
863 tipc_nlist_init(&dsts, tipc_own_addr(net));
864 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
865 if (!dsts.local && !dsts.remote)
866 return -EHOSTUNREACH;
868 /* Build message header */
869 msg_set_type(hdr, TIPC_MCAST_MSG);
870 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
871 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
872 msg_set_destport(hdr, 0);
873 msg_set_destnode(hdr, 0);
874 msg_set_nametype(hdr, ua->sr.type);
875 msg_set_namelower(hdr, ua->sr.lower);
876 msg_set_nameupper(hdr, ua->sr.upper);
878 /* Build message as chain of buffers */
879 __skb_queue_head_init(&pkts);
880 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
882 /* Send message if build was successful */
883 if (unlikely(rc == dlen)) {
884 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
885 TIPC_DUMP_SK_SNDQ, " ");
886 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
887 &tsk->cong_link_cnt);
890 tipc_nlist_purge(&dsts);
892 return rc ? rc : dlen;
896 * tipc_send_group_msg - send a message to a member in the group
897 * @net: network namespace
899 * @m: message to send
901 * @dnode: destination node
902 * @dport: destination port
903 * @dlen: total length of message data
905 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
906 struct msghdr *m, struct tipc_member *mb,
907 u32 dnode, u32 dport, int dlen)
909 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
910 struct tipc_mc_method *method = &tsk->mc_method;
911 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
912 struct tipc_msg *hdr = &tsk->phdr;
913 struct sk_buff_head pkts;
916 /* Complete message header */
917 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
918 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
919 msg_set_destport(hdr, dport);
920 msg_set_destnode(hdr, dnode);
921 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
923 /* Build message as chain of buffers */
924 __skb_queue_head_init(&pkts);
925 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
926 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
927 if (unlikely(rc != dlen))
931 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
932 if (unlikely(rc == -ELINKCONG)) {
933 tipc_dest_push(&tsk->cong_links, dnode, 0);
934 tsk->cong_link_cnt++;
937 /* Update send window */
938 tipc_group_update_member(mb, blks);
940 /* A broadcast sent within next EXPIRE period must follow same path */
941 method->rcast = true;
942 method->mandatory = true;
947 * tipc_send_group_unicast - send message to a member in the group
948 * @sock: socket structure
949 * @m: message to send
950 * @dlen: total length of message data
951 * @timeout: timeout to wait for wakeup
953 * Called from function tipc_sendmsg(), which has done all sanity checks
954 * Return: the number of bytes sent on success, or errno
956 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
957 int dlen, long timeout)
959 struct sock *sk = sock->sk;
960 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
961 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
962 struct tipc_sock *tsk = tipc_sk(sk);
963 struct net *net = sock_net(sk);
964 struct tipc_member *mb = NULL;
971 return -EHOSTUNREACH;
973 /* Block or return if destination link or member is congested */
974 rc = tipc_wait_for_cond(sock, &timeout,
975 !tipc_dest_find(&tsk->cong_links, node, 0) &&
977 !tipc_group_cong(tsk->group, node, port, blks,
983 return -EHOSTUNREACH;
985 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
987 return rc ? rc : dlen;
991 * tipc_send_group_anycast - send message to any member with given identity
992 * @sock: socket structure
993 * @m: message to send
994 * @dlen: total length of message data
995 * @timeout: timeout to wait for wakeup
997 * Called from function tipc_sendmsg(), which has done all sanity checks
998 * Return: the number of bytes sent on success, or errno
1000 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1001 int dlen, long timeout)
1003 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1004 struct sock *sk = sock->sk;
1005 struct tipc_sock *tsk = tipc_sk(sk);
1006 struct list_head *cong_links = &tsk->cong_links;
1007 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1008 struct tipc_msg *hdr = &tsk->phdr;
1009 struct tipc_member *first = NULL;
1010 struct tipc_member *mbr = NULL;
1011 struct net *net = sock_net(sk);
1012 u32 node, port, exclude;
1013 struct list_head dsts;
1018 INIT_LIST_HEAD(&dsts);
1019 ua->sa.type = msg_nametype(hdr);
1020 ua->scope = msg_lookup_scope(hdr);
1022 while (++lookups < 4) {
1023 exclude = tipc_group_exclude(tsk->group);
1027 /* Look for a non-congested destination member, if any */
1029 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1031 return -EHOSTUNREACH;
1032 tipc_dest_pop(&dsts, &node, &port);
1033 cong = tipc_group_cong(tsk->group, node, port, blks,
1043 /* Start over if destination was not in member list */
1047 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1050 /* Block or return if destination link or member is congested */
1051 rc = tipc_wait_for_cond(sock, &timeout,
1052 !tipc_dest_find(cong_links, node, 0) &&
1054 !tipc_group_cong(tsk->group, node, port,
1059 /* Send, unless destination disappeared while waiting */
1064 if (unlikely(lookups >= 4))
1065 return -EHOSTUNREACH;
1067 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1069 return rc ? rc : dlen;
1073 * tipc_send_group_bcast - send message to all members in communication group
1074 * @sock: socket structure
1075 * @m: message to send
1076 * @dlen: total length of message data
1077 * @timeout: timeout to wait for wakeup
1079 * Called from function tipc_sendmsg(), which has done all sanity checks
1080 * Return: the number of bytes sent on success, or errno
1082 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1083 int dlen, long timeout)
1085 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1086 struct sock *sk = sock->sk;
1087 struct net *net = sock_net(sk);
1088 struct tipc_sock *tsk = tipc_sk(sk);
1089 struct tipc_nlist *dsts;
1090 struct tipc_mc_method *method = &tsk->mc_method;
1091 bool ack = method->mandatory && method->rcast;
1092 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1093 struct tipc_msg *hdr = &tsk->phdr;
1094 int mtu = tipc_bcast_get_mtu(net);
1095 struct sk_buff_head pkts;
1096 int rc = -EHOSTUNREACH;
1098 /* Block or return if any destination link or member is congested */
1099 rc = tipc_wait_for_cond(sock, &timeout,
1100 !tsk->cong_link_cnt && tsk->group &&
1101 !tipc_group_bc_cong(tsk->group, blks));
1105 dsts = tipc_group_dests(tsk->group);
1106 if (!dsts->local && !dsts->remote)
1107 return -EHOSTUNREACH;
1109 /* Complete message header */
1111 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1112 msg_set_nameinst(hdr, ua->sa.instance);
1114 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1115 msg_set_nameinst(hdr, 0);
1117 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1118 msg_set_destport(hdr, 0);
1119 msg_set_destnode(hdr, 0);
1120 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1122 /* Avoid getting stuck with repeated forced replicasts */
1123 msg_set_grp_bc_ack_req(hdr, ack);
1125 /* Build message as chain of buffers */
1126 __skb_queue_head_init(&pkts);
1127 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1128 if (unlikely(rc != dlen))
1132 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1136 /* Update broadcast sequence number and send windows */
1137 tipc_group_update_bc_members(tsk->group, blks, ack);
1139 /* Broadcast link is now free to choose method for next broadcast */
1140 method->mandatory = false;
1141 method->expires = jiffies;
1147 * tipc_send_group_mcast - send message to all members with given identity
1148 * @sock: socket structure
1149 * @m: message to send
1150 * @dlen: total length of message data
1151 * @timeout: timeout to wait for wakeup
1153 * Called from function tipc_sendmsg(), which has done all sanity checks
1154 * Return: the number of bytes sent on success, or errno
1156 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1157 int dlen, long timeout)
1159 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1160 struct sock *sk = sock->sk;
1161 struct tipc_sock *tsk = tipc_sk(sk);
1162 struct tipc_group *grp = tsk->group;
1163 struct tipc_msg *hdr = &tsk->phdr;
1164 struct net *net = sock_net(sk);
1165 struct list_head dsts;
1166 u32 dstcnt, exclude;
1168 INIT_LIST_HEAD(&dsts);
1169 ua->sa.type = msg_nametype(hdr);
1170 ua->scope = msg_lookup_scope(hdr);
1171 exclude = tipc_group_exclude(grp);
1173 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1174 return -EHOSTUNREACH;
1177 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1178 return tipc_send_group_unicast(sock, m, dlen, timeout);
1181 tipc_dest_list_purge(&dsts);
1182 return tipc_send_group_bcast(sock, m, dlen, timeout);
1186 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1187 * @net: the associated network namespace
1188 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1189 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1191 * Multi-threaded: parallel calls with reference to same queues may occur
1193 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1194 struct sk_buff_head *inputq)
1196 u32 self = tipc_own_addr(net);
1197 struct sk_buff *skb, *_skb;
1199 struct sk_buff_head tmpq;
1200 struct list_head dports;
1201 struct tipc_msg *hdr;
1202 struct tipc_uaddr ua;
1203 int user, mtyp, hlen;
1205 __skb_queue_head_init(&tmpq);
1206 INIT_LIST_HEAD(&dports);
1207 ua.addrtype = TIPC_SERVICE_RANGE;
1209 /* tipc_skb_peek() increments the head skb's reference counter */
1210 skb = tipc_skb_peek(arrvq, &inputq->lock);
1211 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1213 user = msg_user(hdr);
1214 mtyp = msg_type(hdr);
1215 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1216 onode = msg_orignode(hdr);
1217 ua.sr.type = msg_nametype(hdr);
1218 ua.sr.lower = msg_namelower(hdr);
1219 ua.sr.upper = msg_nameupper(hdr);
1221 ua.scope = TIPC_ANY_SCOPE;
1223 ua.scope = TIPC_CLUSTER_SCOPE;
1225 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1226 spin_lock_bh(&inputq->lock);
1227 if (skb_peek(arrvq) == skb) {
1228 __skb_dequeue(arrvq);
1229 __skb_queue_tail(inputq, skb);
1232 spin_unlock_bh(&inputq->lock);
1236 /* Group messages require exact scope match */
1237 if (msg_in_group(hdr)) {
1240 ua.scope = msg_lookup_scope(hdr);
1243 /* Create destination port list: */
1244 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1246 /* Clone message per destination */
1247 while (tipc_dest_pop(&dports, NULL, &portid)) {
1248 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1250 msg_set_destport(buf_msg(_skb), portid);
1251 __skb_queue_tail(&tmpq, _skb);
1254 pr_warn("Failed to clone mcast rcv buffer\n");
1256 /* Append clones to inputq only if skb is still head of arrvq */
1257 spin_lock_bh(&inputq->lock);
1258 if (skb_peek(arrvq) == skb) {
1259 skb_queue_splice_tail_init(&tmpq, inputq);
1260 /* Decrement the skb's refcnt */
1261 kfree_skb(__skb_dequeue(arrvq));
1263 spin_unlock_bh(&inputq->lock);
1264 __skb_queue_purge(&tmpq);
1267 tipc_sk_rcv(net, inputq);
1270 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1271 * when socket is in Nagle mode
1273 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1275 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1276 struct sk_buff *skb = skb_peek_tail(txq);
1277 struct net *net = sock_net(&tsk->sk);
1278 u32 dnode = tsk_peer_node(tsk);
1282 tsk->pkt_cnt += skb_queue_len(txq);
1283 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1285 if (tsk->nagle_start < NAGLE_START_MAX)
1286 tsk->nagle_start *= 2;
1287 tsk->expect_ack = false;
1288 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1289 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1292 tsk->nagle_start = NAGLE_START_INIT;
1294 msg_set_ack_required(buf_msg(skb));
1295 tsk->expect_ack = true;
1297 tsk->expect_ack = false;
1304 if (!skb || tsk->cong_link_cnt)
1307 /* Do not send SYN again after congestion */
1308 if (msg_is_syn(buf_msg(skb)))
1312 tsk->pkt_cnt += skb_queue_len(txq);
1313 tsk->snt_unacked += tsk->snd_backlog;
1314 tsk->snd_backlog = 0;
1315 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1316 if (rc == -ELINKCONG)
1317 tsk->cong_link_cnt = 1;
1321 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1322 * @tsk: receiving socket
1323 * @skb: pointer to message buffer.
1324 * @inputq: buffer list containing the buffers
1325 * @xmitq: output message area
1327 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1328 struct sk_buff_head *inputq,
1329 struct sk_buff_head *xmitq)
1331 struct tipc_msg *hdr = buf_msg(skb);
1332 u32 onode = tsk_own_node(tsk);
1333 struct sock *sk = &tsk->sk;
1334 int mtyp = msg_type(hdr);
1337 /* Ignore if connection cannot be validated: */
1338 if (!tsk_peer_msg(tsk, hdr)) {
1339 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1343 if (unlikely(msg_errcode(hdr))) {
1344 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1345 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1346 tsk_peer_port(tsk));
1347 sk->sk_state_change(sk);
1349 /* State change is ignored if socket already awake,
1350 * - convert msg to abort msg and add to inqueue
1352 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1353 msg_set_type(hdr, TIPC_CONN_MSG);
1354 msg_set_size(hdr, BASIC_H_SIZE);
1355 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1356 __skb_queue_tail(inputq, skb);
1360 tsk->probe_unacked = false;
1362 if (mtyp == CONN_PROBE) {
1363 msg_set_type(hdr, CONN_PROBE_REPLY);
1364 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1365 __skb_queue_tail(xmitq, skb);
1367 } else if (mtyp == CONN_ACK) {
1368 was_cong = tsk_conn_cong(tsk);
1369 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1370 tsk->snt_unacked -= msg_conn_ack(hdr);
1371 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1372 tsk->snd_win = msg_adv_win(hdr);
1373 if (was_cong && !tsk_conn_cong(tsk))
1374 sk->sk_write_space(sk);
1375 } else if (mtyp != CONN_PROBE_REPLY) {
1376 pr_warn("Received unknown CONN_PROTO msg\n");
1383 * tipc_sendmsg - send message in connectionless manner
1384 * @sock: socket structure
1385 * @m: message to send
1386 * @dsz: amount of user data to be sent
1388 * Message must have an destination specified explicitly.
1389 * Used for SOCK_RDM and SOCK_DGRAM messages,
1390 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1391 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1393 * Return: the number of bytes sent on success, or errno otherwise
1395 static int tipc_sendmsg(struct socket *sock,
1396 struct msghdr *m, size_t dsz)
1398 struct sock *sk = sock->sk;
1402 ret = __tipc_sendmsg(sock, m, dsz);
1408 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1410 struct sock *sk = sock->sk;
1411 struct net *net = sock_net(sk);
1412 struct tipc_sock *tsk = tipc_sk(sk);
1413 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1414 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1415 struct list_head *clinks = &tsk->cong_links;
1416 bool syn = !tipc_sk_type_connectionless(sk);
1417 struct tipc_group *grp = tsk->group;
1418 struct tipc_msg *hdr = &tsk->phdr;
1419 struct tipc_socket_addr skaddr;
1420 struct sk_buff_head pkts;
1423 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1427 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1429 atype = ua->addrtype;
1432 /* If socket belongs to a communication group follow other paths */
1435 return tipc_send_group_bcast(sock, m, dlen, timeout);
1436 if (atype == TIPC_SERVICE_ADDR)
1437 return tipc_send_group_anycast(sock, m, dlen, timeout);
1438 if (atype == TIPC_SOCKET_ADDR)
1439 return tipc_send_group_unicast(sock, m, dlen, timeout);
1440 if (atype == TIPC_SERVICE_RANGE)
1441 return tipc_send_group_mcast(sock, m, dlen, timeout);
1446 ua = (struct tipc_uaddr *)&tsk->peer;
1447 if (!syn && ua->family != AF_TIPC)
1448 return -EDESTADDRREQ;
1449 atype = ua->addrtype;
1452 if (unlikely(syn)) {
1453 if (sk->sk_state == TIPC_LISTEN)
1455 if (sk->sk_state != TIPC_OPEN)
1459 if (atype == TIPC_SERVICE_ADDR)
1460 tsk->conn_addrtype = atype;
1461 msg_set_syn(hdr, 1);
1464 memset(&skaddr, 0, sizeof(skaddr));
1466 /* Determine destination */
1467 if (atype == TIPC_SERVICE_RANGE) {
1468 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1469 } else if (atype == TIPC_SERVICE_ADDR) {
1470 skaddr.node = ua->lookup_node;
1471 ua->scope = tipc_node2scope(skaddr.node);
1472 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1473 return -EHOSTUNREACH;
1474 } else if (atype == TIPC_SOCKET_ADDR) {
1480 /* Block or return if destination link is congested */
1481 rc = tipc_wait_for_cond(sock, &timeout,
1482 !tipc_dest_find(clinks, skaddr.node, 0));
1486 /* Finally build message header */
1487 msg_set_destnode(hdr, skaddr.node);
1488 msg_set_destport(hdr, skaddr.ref);
1489 if (atype == TIPC_SERVICE_ADDR) {
1490 msg_set_type(hdr, TIPC_NAMED_MSG);
1491 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1492 msg_set_nametype(hdr, ua->sa.type);
1493 msg_set_nameinst(hdr, ua->sa.instance);
1494 msg_set_lookup_scope(hdr, ua->scope);
1495 } else { /* TIPC_SOCKET_ADDR */
1496 msg_set_type(hdr, TIPC_DIRECT_MSG);
1497 msg_set_lookup_scope(hdr, 0);
1498 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1501 /* Add message body */
1502 __skb_queue_head_init(&pkts);
1503 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1504 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1505 if (unlikely(rc != dlen))
1507 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1508 __skb_queue_purge(&pkts);
1513 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1514 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1515 if (unlikely(rc == -ELINKCONG)) {
1516 tipc_dest_push(clinks, skaddr.node, 0);
1517 tsk->cong_link_cnt++;
1521 if (unlikely(syn && !rc)) {
1522 tipc_set_sk_state(sk, TIPC_CONNECTING);
1523 if (dlen && timeout) {
1524 timeout = msecs_to_jiffies(timeout);
1525 tipc_wait_for_connect(sock, &timeout);
1529 return rc ? rc : dlen;
1533 * tipc_sendstream - send stream-oriented data
1534 * @sock: socket structure
1536 * @dsz: total length of data to be transmitted
1538 * Used for SOCK_STREAM data.
1540 * Return: the number of bytes sent on success (or partial success),
1541 * or errno if no data sent
1543 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1545 struct sock *sk = sock->sk;
1549 ret = __tipc_sendstream(sock, m, dsz);
1555 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1557 struct sock *sk = sock->sk;
1558 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1559 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1560 struct sk_buff_head *txq = &sk->sk_write_queue;
1561 struct tipc_sock *tsk = tipc_sk(sk);
1562 struct tipc_msg *hdr = &tsk->phdr;
1563 struct net *net = sock_net(sk);
1564 struct sk_buff *skb;
1565 u32 dnode = tsk_peer_node(tsk);
1566 int maxnagle = tsk->maxnagle;
1567 int maxpkt = tsk->max_pkt;
1571 if (unlikely(dlen > INT_MAX))
1574 /* Handle implicit connection setup */
1575 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1576 rc = __tipc_sendmsg(sock, m, dlen);
1577 if (dlen && dlen == rc) {
1578 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1579 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1585 rc = tipc_wait_for_cond(sock, &timeout,
1586 (!tsk->cong_link_cnt &&
1587 !tsk_conn_cong(tsk) &&
1588 tipc_sk_connected(sk)));
1591 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1592 blocks = tsk->snd_backlog;
1593 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1595 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1596 if (unlikely(rc < 0))
1600 if (blocks <= 64 && tsk->expect_ack) {
1601 tsk->snd_backlog = blocks;
1604 } else if (blocks > 64) {
1605 tsk->pkt_cnt += skb_queue_len(txq);
1607 skb = skb_peek_tail(txq);
1609 msg_set_ack_required(buf_msg(skb));
1610 tsk->expect_ack = true;
1612 tsk->expect_ack = false;
1618 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1619 if (unlikely(rc != send))
1621 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1623 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1624 TIPC_DUMP_SK_SNDQ, " ");
1625 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1626 if (unlikely(rc == -ELINKCONG)) {
1627 tsk->cong_link_cnt = 1;
1631 tsk->snt_unacked += blocks;
1632 tsk->snd_backlog = 0;
1635 } while (sent < dlen && !rc);
1637 return sent ? sent : rc;
1641 * tipc_send_packet - send a connection-oriented message
1642 * @sock: socket structure
1643 * @m: message to send
1644 * @dsz: length of data to be transmitted
1646 * Used for SOCK_SEQPACKET messages.
1648 * Return: the number of bytes sent on success, or errno otherwise
1650 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1652 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1655 return tipc_sendstream(sock, m, dsz);
1658 /* tipc_sk_finish_conn - complete the setup of a connection
1660 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1663 struct sock *sk = &tsk->sk;
1664 struct net *net = sock_net(sk);
1665 struct tipc_msg *msg = &tsk->phdr;
1667 msg_set_syn(msg, 0);
1668 msg_set_destnode(msg, peer_node);
1669 msg_set_destport(msg, peer_port);
1670 msg_set_type(msg, TIPC_CONN_MSG);
1671 msg_set_lookup_scope(msg, 0);
1672 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1674 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1675 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1676 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1677 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1678 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1680 __skb_queue_purge(&sk->sk_write_queue);
1681 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1684 /* Fall back to message based flow control */
1685 tsk->rcv_win = FLOWCTL_MSG_WIN;
1686 tsk->snd_win = FLOWCTL_MSG_WIN;
1690 * tipc_sk_set_orig_addr - capture sender's address for received message
1691 * @m: descriptor for message info
1692 * @skb: received message
1694 * Note: Address is not captured if not requested by receiver.
1696 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1698 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1699 struct tipc_msg *hdr = buf_msg(skb);
1704 srcaddr->sock.family = AF_TIPC;
1705 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1706 srcaddr->sock.scope = 0;
1707 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1708 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1709 srcaddr->sock.addr.name.domain = 0;
1710 m->msg_namelen = sizeof(struct sockaddr_tipc);
1712 if (!msg_in_group(hdr))
1715 /* Group message users may also want to know sending member's id */
1716 srcaddr->member.family = AF_TIPC;
1717 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1718 srcaddr->member.scope = 0;
1719 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1720 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1721 srcaddr->member.addr.name.domain = 0;
1722 m->msg_namelen = sizeof(*srcaddr);
1726 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1727 * @m: descriptor for message info
1728 * @skb: received message buffer
1729 * @tsk: TIPC port associated with message
1731 * Note: Ancillary data is not captured if not requested by receiver.
1733 * Return: 0 if successful, otherwise errno
1735 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1736 struct tipc_sock *tsk)
1738 struct tipc_msg *hdr;
1743 if (likely(m->msg_controllen == 0))
1747 dlen = msg_data_sz(hdr);
1749 /* Capture errored message object, if any */
1750 if (msg_errcode(hdr)) {
1751 if (skb_linearize(skb))
1754 data[0] = msg_errcode(hdr);
1756 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1759 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1764 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1765 switch (msg_type(hdr)) {
1766 case TIPC_NAMED_MSG:
1768 data[0] = msg_nametype(hdr);
1769 data[1] = msg_namelower(hdr);
1772 case TIPC_MCAST_MSG:
1774 data[0] = msg_nametype(hdr);
1775 data[1] = msg_namelower(hdr);
1776 data[2] = msg_nameupper(hdr);
1779 has_addr = !!tsk->conn_addrtype;
1780 data[0] = msg_nametype(&tsk->phdr);
1781 data[1] = msg_nameinst(&tsk->phdr);
1789 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1792 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1794 struct sock *sk = &tsk->sk;
1795 struct sk_buff *skb = NULL;
1796 struct tipc_msg *msg;
1797 u32 peer_port = tsk_peer_port(tsk);
1798 u32 dnode = tsk_peer_node(tsk);
1800 if (!tipc_sk_connected(sk))
1802 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1803 dnode, tsk_own_node(tsk), peer_port,
1804 tsk->portid, TIPC_OK);
1808 msg_set_conn_ack(msg, tsk->rcv_unacked);
1809 tsk->rcv_unacked = 0;
1811 /* Adjust to and advertize the correct window limit */
1812 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1813 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1814 msg_set_adv_win(msg, tsk->rcv_win);
1819 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1821 struct sk_buff *skb;
1823 skb = tipc_sk_build_ack(tsk);
1827 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1828 msg_link_selector(buf_msg(skb)));
1831 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1833 struct sock *sk = sock->sk;
1834 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1835 long timeo = *timeop;
1836 int err = sock_error(sk);
1842 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1843 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1847 add_wait_queue(sk_sleep(sk), &wait);
1849 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1850 sched_annotate_sleep();
1852 remove_wait_queue(sk_sleep(sk), &wait);
1855 if (!skb_queue_empty(&sk->sk_receive_queue))
1860 err = sock_intr_errno(timeo);
1861 if (signal_pending(current))
1864 err = sock_error(sk);
1873 * tipc_recvmsg - receive packet-oriented message
1874 * @sock: network socket
1875 * @m: descriptor for message info
1876 * @buflen: length of user buffer area
1877 * @flags: receive flags
1879 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1880 * If the complete message doesn't fit in user area, truncate it.
1882 * Return: size of returned message data, errno otherwise
1884 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1885 size_t buflen, int flags)
1887 struct sock *sk = sock->sk;
1888 bool connected = !tipc_sk_type_connectionless(sk);
1889 struct tipc_sock *tsk = tipc_sk(sk);
1890 int rc, err, hlen, dlen, copy;
1891 struct tipc_skb_cb *skb_cb;
1892 struct sk_buff_head xmitq;
1893 struct tipc_msg *hdr;
1894 struct sk_buff *skb;
1898 /* Catch invalid receive requests */
1899 if (unlikely(!buflen))
1903 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1907 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1909 /* Step rcv queue to first msg with data or error; wait if necessary */
1911 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1914 skb = skb_peek(&sk->sk_receive_queue);
1915 skb_cb = TIPC_SKB_CB(skb);
1917 dlen = msg_data_sz(hdr);
1918 hlen = msg_hdr_sz(hdr);
1919 err = msg_errcode(hdr);
1920 grp_evt = msg_is_grp_evt(hdr);
1921 if (likely(dlen || err))
1923 tsk_advance_rx_queue(sk);
1926 /* Collect msg meta data, including error code and rejected data */
1927 tipc_sk_set_orig_addr(m, skb);
1928 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1933 /* Capture data if non-error msg, otherwise just set return value */
1935 int offset = skb_cb->bytes_read;
1937 copy = min_t(int, dlen - offset, buflen);
1938 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1941 if (unlikely(offset + copy < dlen)) {
1942 if (flags & MSG_EOR) {
1943 if (!(flags & MSG_PEEK))
1944 skb_cb->bytes_read = offset + copy;
1946 m->msg_flags |= MSG_TRUNC;
1947 skb_cb->bytes_read = 0;
1950 if (flags & MSG_EOR)
1951 m->msg_flags |= MSG_EOR;
1952 skb_cb->bytes_read = 0;
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 if (skb_cb->bytes_read)
1987 tsk_advance_rx_queue(sk);
1989 if (likely(!connected))
1992 /* Send connection flow control advertisement when applicable */
1993 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1994 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1995 tipc_sk_send_ack(tsk);
1998 return rc ? rc : copy;
2002 * tipc_recvstream - receive stream-oriented data
2003 * @sock: network socket
2004 * @m: descriptor for message info
2005 * @buflen: total size of user buffer area
2006 * @flags: receive flags
2008 * Used for SOCK_STREAM messages only. If not enough data is available
2009 * will optionally wait for more; never truncates data.
2011 * Return: size of returned message data, errno otherwise
2013 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2014 size_t buflen, int flags)
2016 struct sock *sk = sock->sk;
2017 struct tipc_sock *tsk = tipc_sk(sk);
2018 struct sk_buff *skb;
2019 struct tipc_msg *hdr;
2020 struct tipc_skb_cb *skb_cb;
2021 bool peek = flags & MSG_PEEK;
2022 int offset, required, copy, copied = 0;
2023 int hlen, dlen, err, rc;
2026 /* Catch invalid receive attempts */
2027 if (unlikely(!buflen))
2032 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2036 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2037 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2040 /* Look at first msg in receive queue; wait if necessary */
2041 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2044 skb = skb_peek(&sk->sk_receive_queue);
2045 skb_cb = TIPC_SKB_CB(skb);
2047 dlen = msg_data_sz(hdr);
2048 hlen = msg_hdr_sz(hdr);
2049 err = msg_errcode(hdr);
2051 /* Discard any empty non-errored (SYN-) message */
2052 if (unlikely(!dlen && !err)) {
2053 tsk_advance_rx_queue(sk);
2057 /* Collect msg meta data, incl. error code and rejected data */
2059 tipc_sk_set_orig_addr(m, skb);
2060 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2066 /* Copy data if msg ok, otherwise return error/partial data */
2068 offset = skb_cb->bytes_read;
2069 copy = min_t(int, dlen - offset, buflen - copied);
2070 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2075 if (unlikely(offset < dlen)) {
2077 skb_cb->bytes_read = offset;
2082 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2091 tsk_advance_rx_queue(sk);
2093 /* Send connection flow control advertisement when applicable */
2094 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2095 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2096 tipc_sk_send_ack(tsk);
2098 /* Exit if all requested data or FIN/error received */
2099 if (copied == buflen || err)
2102 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2105 return copied ? copied : rc;
2109 * tipc_write_space - wake up thread if port congestion is released
2112 static void tipc_write_space(struct sock *sk)
2114 struct socket_wq *wq;
2117 wq = rcu_dereference(sk->sk_wq);
2118 if (skwq_has_sleeper(wq))
2119 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2120 EPOLLWRNORM | EPOLLWRBAND);
2125 * tipc_data_ready - wake up threads to indicate messages have been received
2128 static void tipc_data_ready(struct sock *sk)
2130 struct socket_wq *wq;
2133 wq = rcu_dereference(sk->sk_wq);
2134 if (skwq_has_sleeper(wq))
2135 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2136 EPOLLRDNORM | EPOLLRDBAND);
2140 static void tipc_sock_destruct(struct sock *sk)
2142 __skb_queue_purge(&sk->sk_receive_queue);
2145 static void tipc_sk_proto_rcv(struct sock *sk,
2146 struct sk_buff_head *inputq,
2147 struct sk_buff_head *xmitq)
2149 struct sk_buff *skb = __skb_dequeue(inputq);
2150 struct tipc_sock *tsk = tipc_sk(sk);
2151 struct tipc_msg *hdr = buf_msg(skb);
2152 struct tipc_group *grp = tsk->group;
2153 bool wakeup = false;
2155 switch (msg_user(hdr)) {
2157 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2160 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2161 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2163 tsk->cong_link_cnt--;
2165 tipc_sk_push_backlog(tsk, false);
2167 case GROUP_PROTOCOL:
2168 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2171 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2172 hdr, inputq, xmitq);
2179 sk->sk_write_space(sk);
2185 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2187 * @skb: pointer to message buffer.
2188 * @xmitq: for Nagle ACK if any
2189 * Return: true if message should be added to receive queue, false otherwise
2191 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2192 struct sk_buff_head *xmitq)
2194 struct sock *sk = &tsk->sk;
2195 struct net *net = sock_net(sk);
2196 struct tipc_msg *hdr = buf_msg(skb);
2197 bool con_msg = msg_connected(hdr);
2198 u32 pport = tsk_peer_port(tsk);
2199 u32 pnode = tsk_peer_node(tsk);
2200 u32 oport = msg_origport(hdr);
2201 u32 onode = msg_orignode(hdr);
2202 int err = msg_errcode(hdr);
2203 unsigned long delay;
2205 if (unlikely(msg_mcast(hdr)))
2209 switch (sk->sk_state) {
2210 case TIPC_CONNECTING:
2212 if (likely(con_msg)) {
2215 tipc_sk_finish_conn(tsk, oport, onode);
2216 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2217 /* ACK+ message with data is added to receive queue */
2218 if (msg_data_sz(hdr))
2220 /* Empty ACK-, - wake up sleeping connect() and drop */
2221 sk->sk_state_change(sk);
2222 msg_set_dest_droppable(hdr, 1);
2225 /* Ignore connectionless message if not from listening socket */
2226 if (oport != pport || onode != pnode)
2230 if (err != TIPC_ERR_OVERLOAD)
2233 /* Prepare for new setup attempt if we have a SYN clone */
2234 if (skb_queue_empty(&sk->sk_write_queue))
2236 get_random_bytes(&delay, 2);
2237 delay %= (tsk->conn_timeout / 4);
2238 delay = msecs_to_jiffies(delay + 100);
2239 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2242 case TIPC_DISCONNECTING:
2245 /* Accept only SYN message */
2246 if (!msg_is_syn(hdr) &&
2247 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2249 if (!con_msg && !err)
2252 case TIPC_ESTABLISHED:
2253 if (!skb_queue_empty(&sk->sk_write_queue))
2254 tipc_sk_push_backlog(tsk, false);
2255 /* Accept only connection-based messages sent by peer */
2256 if (likely(con_msg && !err && pport == oport &&
2258 if (msg_ack_required(hdr)) {
2259 struct sk_buff *skb;
2261 skb = tipc_sk_build_ack(tsk);
2263 msg_set_nagle_ack(buf_msg(skb));
2264 __skb_queue_tail(xmitq, skb);
2269 if (!tsk_peer_msg(tsk, hdr))
2273 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2274 tipc_node_remove_conn(net, pnode, tsk->portid);
2275 sk->sk_state_change(sk);
2278 pr_err("Unknown sk_state %u\n", sk->sk_state);
2280 /* Abort connection setup attempt */
2281 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2282 sk->sk_err = ECONNREFUSED;
2283 sk->sk_state_change(sk);
2288 * rcvbuf_limit - get proper overload limit of socket receive queue
2292 * For connection oriented messages, irrespective of importance,
2293 * default queue limit is 2 MB.
2295 * For connectionless messages, queue limits are based on message
2296 * importance as follows:
2298 * TIPC_LOW_IMPORTANCE (2 MB)
2299 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2300 * TIPC_HIGH_IMPORTANCE (8 MB)
2301 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2303 * Return: overload limit according to corresponding message importance
2305 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2307 struct tipc_sock *tsk = tipc_sk(sk);
2308 struct tipc_msg *hdr = buf_msg(skb);
2310 if (unlikely(msg_in_group(hdr)))
2311 return READ_ONCE(sk->sk_rcvbuf);
2313 if (unlikely(!msg_connected(hdr)))
2314 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2316 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2317 return READ_ONCE(sk->sk_rcvbuf);
2319 return FLOWCTL_MSG_LIM;
2323 * tipc_sk_filter_rcv - validate incoming message
2325 * @skb: pointer to message.
2326 * @xmitq: output message area (FIXME)
2328 * Enqueues message on receive queue if acceptable; optionally handles
2329 * disconnect indication for a connected socket.
2331 * Called with socket lock already taken
2333 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2334 struct sk_buff_head *xmitq)
2336 bool sk_conn = !tipc_sk_type_connectionless(sk);
2337 struct tipc_sock *tsk = tipc_sk(sk);
2338 struct tipc_group *grp = tsk->group;
2339 struct tipc_msg *hdr = buf_msg(skb);
2340 struct net *net = sock_net(sk);
2341 struct sk_buff_head inputq;
2342 int mtyp = msg_type(hdr);
2343 int limit, err = TIPC_OK;
2345 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2346 TIPC_SKB_CB(skb)->bytes_read = 0;
2347 __skb_queue_head_init(&inputq);
2348 __skb_queue_tail(&inputq, skb);
2350 if (unlikely(!msg_isdata(hdr)))
2351 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2354 tipc_group_filter_msg(grp, &inputq, xmitq);
2356 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2357 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2359 /* Validate and add to receive buffer if there is space */
2360 while ((skb = __skb_dequeue(&inputq))) {
2362 limit = rcvbuf_limit(sk, skb);
2363 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2364 (!sk_conn && msg_connected(hdr)) ||
2365 (!grp && msg_in_group(hdr)))
2366 err = TIPC_ERR_NO_PORT;
2367 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2368 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2370 atomic_inc(&sk->sk_drops);
2371 err = TIPC_ERR_OVERLOAD;
2374 if (unlikely(err)) {
2375 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2376 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2378 __skb_queue_tail(xmitq, skb);
2383 __skb_queue_tail(&sk->sk_receive_queue, skb);
2384 skb_set_owner_r(skb, sk);
2385 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2386 "rcvq >90% allocated!");
2387 sk->sk_data_ready(sk);
2392 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2396 * Caller must hold socket lock
2398 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2400 unsigned int before = sk_rmem_alloc_get(sk);
2401 struct sk_buff_head xmitq;
2404 __skb_queue_head_init(&xmitq);
2406 tipc_sk_filter_rcv(sk, skb, &xmitq);
2407 added = sk_rmem_alloc_get(sk) - before;
2408 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2410 /* Send pending response/rejected messages, if any */
2411 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2416 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2417 * inputq and try adding them to socket or backlog queue
2418 * @inputq: list of incoming buffers with potentially different destinations
2419 * @sk: socket where the buffers should be enqueued
2420 * @dport: port number for the socket
2421 * @xmitq: output queue
2423 * Caller must hold socket lock
2425 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2426 u32 dport, struct sk_buff_head *xmitq)
2428 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2429 struct sk_buff *skb;
2434 while (skb_queue_len(inputq)) {
2435 if (unlikely(time_after_eq(jiffies, time_limit)))
2438 skb = tipc_skb_dequeue(inputq, dport);
2442 /* Add message directly to receive queue if possible */
2443 if (!sock_owned_by_user(sk)) {
2444 tipc_sk_filter_rcv(sk, skb, xmitq);
2448 /* Try backlog, compensating for double-counted bytes */
2449 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2450 if (!sk->sk_backlog.len)
2451 atomic_set(dcnt, 0);
2452 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2453 if (likely(!sk_add_backlog(sk, skb, lim))) {
2454 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2455 "bklg & rcvq >90% allocated!");
2459 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2460 /* Overload => reject message back to sender */
2461 onode = tipc_own_addr(sock_net(sk));
2462 atomic_inc(&sk->sk_drops);
2463 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2464 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2466 __skb_queue_tail(xmitq, skb);
2473 * tipc_sk_rcv - handle a chain of incoming buffers
2474 * @net: the associated network namespace
2475 * @inputq: buffer list containing the buffers
2476 * Consumes all buffers in list until inputq is empty
2477 * Note: may be called in multiple threads referring to the same queue
2479 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2481 struct sk_buff_head xmitq;
2482 u32 dnode, dport = 0;
2484 struct tipc_sock *tsk;
2486 struct sk_buff *skb;
2488 __skb_queue_head_init(&xmitq);
2489 while (skb_queue_len(inputq)) {
2490 dport = tipc_skb_peek_port(inputq, dport);
2491 tsk = tipc_sk_lookup(net, dport);
2495 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2496 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2497 spin_unlock_bh(&sk->sk_lock.slock);
2499 /* Send pending response/rejected messages, if any */
2500 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2504 /* No destination socket => dequeue skb if still there */
2505 skb = tipc_skb_dequeue(inputq, dport);
2509 /* Try secondary lookup if unresolved named message */
2510 err = TIPC_ERR_NO_PORT;
2511 if (tipc_msg_lookup_dest(net, skb, &err))
2514 /* Prepare for message rejection */
2515 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2518 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2520 dnode = msg_destnode(buf_msg(skb));
2521 tipc_node_xmit_skb(net, skb, dnode, dport);
2525 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2527 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2528 struct sock *sk = sock->sk;
2532 int err = sock_error(sk);
2537 if (signal_pending(current))
2538 return sock_intr_errno(*timeo_p);
2539 if (sk->sk_state == TIPC_DISCONNECTING)
2542 add_wait_queue(sk_sleep(sk), &wait);
2543 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2545 remove_wait_queue(sk_sleep(sk), &wait);
2550 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2552 if (addr->family != AF_TIPC)
2554 if (addr->addrtype == TIPC_SERVICE_RANGE)
2555 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2556 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2557 addr->addrtype == TIPC_SOCKET_ADDR);
2561 * tipc_connect - establish a connection to another TIPC port
2562 * @sock: socket structure
2563 * @dest: socket address for destination port
2564 * @destlen: size of socket address data structure
2565 * @flags: file-related flags associated with socket
2567 * Return: 0 on success, errno otherwise
2569 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2570 int destlen, int flags)
2572 struct sock *sk = sock->sk;
2573 struct tipc_sock *tsk = tipc_sk(sk);
2574 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2575 struct msghdr m = {NULL,};
2576 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2580 if (destlen != sizeof(struct sockaddr_tipc))
2590 if (dst->family == AF_UNSPEC) {
2591 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2592 if (!tipc_sk_type_connectionless(sk))
2596 if (!tipc_sockaddr_is_sane(dst)) {
2600 /* DGRAM/RDM connect(), just save the destaddr */
2601 if (tipc_sk_type_connectionless(sk)) {
2602 memcpy(&tsk->peer, dest, destlen);
2604 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2609 previous = sk->sk_state;
2611 switch (sk->sk_state) {
2613 /* Send a 'SYN-' to destination */
2615 m.msg_namelen = destlen;
2617 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2618 * indicate send_msg() is never blocked.
2621 m.msg_flags = MSG_DONTWAIT;
2623 res = __tipc_sendmsg(sock, &m, 0);
2624 if ((res < 0) && (res != -EWOULDBLOCK))
2627 /* Just entered TIPC_CONNECTING state; the only
2628 * difference is that return value in non-blocking
2629 * case is EINPROGRESS, rather than EALREADY.
2633 case TIPC_CONNECTING:
2635 if (previous == TIPC_CONNECTING)
2639 timeout = msecs_to_jiffies(timeout);
2640 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2641 res = tipc_wait_for_connect(sock, &timeout);
2643 case TIPC_ESTABLISHED:
2656 * tipc_listen - allow socket to listen for incoming connections
2657 * @sock: socket structure
2660 * Return: 0 on success, errno otherwise
2662 static int tipc_listen(struct socket *sock, int len)
2664 struct sock *sk = sock->sk;
2668 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2674 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2676 struct sock *sk = sock->sk;
2677 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2680 /* True wake-one mechanism for incoming connections: only
2681 * one process gets woken up, not the 'whole herd'.
2682 * Since we do not 'race & poll' for established sockets
2683 * anymore, the common case will execute the loop only once.
2686 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2687 add_wait_queue(sk_sleep(sk), &wait);
2689 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2691 remove_wait_queue(sk_sleep(sk), &wait);
2694 if (!skb_queue_empty(&sk->sk_receive_queue))
2699 err = sock_intr_errno(timeo);
2700 if (signal_pending(current))
2707 * tipc_accept - wait for connection request
2708 * @sock: listening socket
2709 * @new_sock: new socket that is to be connected
2710 * @flags: file-related flags associated with socket
2711 * @kern: caused by kernel or by userspace?
2713 * Return: 0 on success, errno otherwise
2715 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2718 struct sock *new_sk, *sk = sock->sk;
2719 struct tipc_sock *new_tsock;
2720 struct msghdr m = {NULL,};
2721 struct tipc_msg *msg;
2722 struct sk_buff *buf;
2728 if (sk->sk_state != TIPC_LISTEN) {
2732 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2733 res = tipc_wait_for_accept(sock, timeo);
2737 buf = skb_peek(&sk->sk_receive_queue);
2739 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2742 security_sk_clone(sock->sk, new_sock->sk);
2744 new_sk = new_sock->sk;
2745 new_tsock = tipc_sk(new_sk);
2748 /* we lock on new_sk; but lockdep sees the lock on sk */
2749 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2752 * Reject any stray messages received by new socket
2753 * before the socket lock was taken (very, very unlikely)
2755 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2757 /* Connect new socket to it's peer */
2758 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2760 tsk_set_importance(new_sk, msg_importance(msg));
2761 if (msg_named(msg)) {
2762 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2763 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2764 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2768 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2769 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2771 if (!msg_data_sz(msg)) {
2772 tsk_advance_rx_queue(sk);
2774 __skb_dequeue(&sk->sk_receive_queue);
2775 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2776 skb_set_owner_r(buf, new_sk);
2778 __tipc_sendstream(new_sock, &m, 0);
2779 release_sock(new_sk);
2786 * tipc_shutdown - shutdown socket connection
2787 * @sock: socket structure
2788 * @how: direction to close (must be SHUT_RDWR)
2790 * Terminates connection (if necessary), then purges socket's receive queue.
2792 * Return: 0 on success, errno otherwise
2794 static int tipc_shutdown(struct socket *sock, int how)
2796 struct sock *sk = sock->sk;
2799 if (how != SHUT_RDWR)
2804 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2805 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2806 sk->sk_shutdown = SHUTDOWN_MASK;
2808 if (sk->sk_state == TIPC_DISCONNECTING) {
2809 /* Discard any unreceived messages */
2810 __skb_queue_purge(&sk->sk_receive_queue);
2816 /* Wake up anyone sleeping in poll. */
2817 sk->sk_state_change(sk);
2823 static void tipc_sk_check_probing_state(struct sock *sk,
2824 struct sk_buff_head *list)
2826 struct tipc_sock *tsk = tipc_sk(sk);
2827 u32 pnode = tsk_peer_node(tsk);
2828 u32 pport = tsk_peer_port(tsk);
2829 u32 self = tsk_own_node(tsk);
2830 u32 oport = tsk->portid;
2831 struct sk_buff *skb;
2833 if (tsk->probe_unacked) {
2834 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2835 sk->sk_err = ECONNABORTED;
2836 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2837 sk->sk_state_change(sk);
2840 /* Prepare new probe */
2841 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2842 pnode, self, pport, oport, TIPC_OK);
2844 __skb_queue_tail(list, skb);
2845 tsk->probe_unacked = true;
2846 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2849 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2851 struct tipc_sock *tsk = tipc_sk(sk);
2853 /* Try again later if dest link is congested */
2854 if (tsk->cong_link_cnt) {
2855 sk_reset_timer(sk, &sk->sk_timer,
2856 jiffies + msecs_to_jiffies(100));
2859 /* Prepare SYN for retransmit */
2860 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2863 static void tipc_sk_timeout(struct timer_list *t)
2865 struct sock *sk = from_timer(sk, t, sk_timer);
2866 struct tipc_sock *tsk = tipc_sk(sk);
2867 u32 pnode = tsk_peer_node(tsk);
2868 struct sk_buff_head list;
2871 __skb_queue_head_init(&list);
2874 /* Try again later if socket is busy */
2875 if (sock_owned_by_user(sk)) {
2876 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2882 if (sk->sk_state == TIPC_ESTABLISHED)
2883 tipc_sk_check_probing_state(sk, &list);
2884 else if (sk->sk_state == TIPC_CONNECTING)
2885 tipc_sk_retry_connect(sk, &list);
2889 if (!skb_queue_empty(&list))
2890 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2892 /* SYN messages may cause link congestion */
2893 if (rc == -ELINKCONG) {
2894 tipc_dest_push(&tsk->cong_links, pnode, 0);
2895 tsk->cong_link_cnt = 1;
2900 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2902 struct sock *sk = &tsk->sk;
2903 struct net *net = sock_net(sk);
2904 struct tipc_socket_addr skaddr;
2905 struct publication *p;
2908 if (tipc_sk_connected(sk))
2910 key = tsk->portid + tsk->pub_count + 1;
2911 if (key == tsk->portid)
2913 skaddr.ref = tsk->portid;
2914 skaddr.node = tipc_own_addr(net);
2915 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2919 list_add(&p->binding_sock, &tsk->publications);
2921 tsk->published = true;
2925 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2927 struct net *net = sock_net(&tsk->sk);
2928 struct publication *safe, *p;
2929 struct tipc_uaddr _ua;
2932 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2934 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2935 p->sr.type, p->sr.lower, p->sr.upper);
2936 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2939 /* Unbind specific publication */
2940 if (p->scope != ua->scope)
2942 if (p->sr.type != ua->sr.type)
2944 if (p->sr.lower != ua->sr.lower)
2946 if (p->sr.upper != ua->sr.upper)
2948 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2952 if (list_empty(&tsk->publications)) {
2959 /* tipc_sk_reinit: set non-zero address in all existing sockets
2960 * when we go from standalone to network mode.
2962 void tipc_sk_reinit(struct net *net)
2964 struct tipc_net *tn = net_generic(net, tipc_net_id);
2965 struct rhashtable_iter iter;
2966 struct tipc_sock *tsk;
2967 struct tipc_msg *msg;
2969 rhashtable_walk_enter(&tn->sk_rht, &iter);
2972 rhashtable_walk_start(&iter);
2974 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2975 sock_hold(&tsk->sk);
2976 rhashtable_walk_stop(&iter);
2977 lock_sock(&tsk->sk);
2979 msg_set_prevnode(msg, tipc_own_addr(net));
2980 msg_set_orignode(msg, tipc_own_addr(net));
2981 release_sock(&tsk->sk);
2982 rhashtable_walk_start(&iter);
2986 rhashtable_walk_stop(&iter);
2987 } while (tsk == ERR_PTR(-EAGAIN));
2989 rhashtable_walk_exit(&iter);
2992 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2994 struct tipc_net *tn = net_generic(net, tipc_net_id);
2995 struct tipc_sock *tsk;
2998 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
3000 sock_hold(&tsk->sk);
3006 static int tipc_sk_insert(struct tipc_sock *tsk)
3008 struct sock *sk = &tsk->sk;
3009 struct net *net = sock_net(sk);
3010 struct tipc_net *tn = net_generic(net, tipc_net_id);
3011 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3012 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3014 while (remaining--) {
3016 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3017 portid = TIPC_MIN_PORT;
3018 tsk->portid = portid;
3019 sock_hold(&tsk->sk);
3020 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3029 static void tipc_sk_remove(struct tipc_sock *tsk)
3031 struct sock *sk = &tsk->sk;
3032 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3034 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3035 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3040 static const struct rhashtable_params tsk_rht_params = {
3042 .head_offset = offsetof(struct tipc_sock, node),
3043 .key_offset = offsetof(struct tipc_sock, portid),
3044 .key_len = sizeof(u32), /* portid */
3045 .max_size = 1048576,
3047 .automatic_shrinking = true,
3050 int tipc_sk_rht_init(struct net *net)
3052 struct tipc_net *tn = net_generic(net, tipc_net_id);
3054 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3057 void tipc_sk_rht_destroy(struct net *net)
3059 struct tipc_net *tn = net_generic(net, tipc_net_id);
3061 /* Wait for socket readers to complete */
3064 rhashtable_destroy(&tn->sk_rht);
3067 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3069 struct net *net = sock_net(&tsk->sk);
3070 struct tipc_group *grp = tsk->group;
3071 struct tipc_msg *hdr = &tsk->phdr;
3072 struct tipc_uaddr ua;
3075 if (mreq->type < TIPC_RESERVED_TYPES)
3077 if (mreq->scope > TIPC_NODE_SCOPE)
3079 if (mreq->scope != TIPC_NODE_SCOPE)
3080 mreq->scope = TIPC_CLUSTER_SCOPE;
3083 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3087 msg_set_lookup_scope(hdr, mreq->scope);
3088 msg_set_nametype(hdr, mreq->type);
3089 msg_set_dest_droppable(hdr, true);
3090 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3091 mreq->type, mreq->instance, mreq->instance);
3092 tipc_nametbl_build_group(net, grp, &ua);
3093 rc = tipc_sk_publish(tsk, &ua);
3095 tipc_group_delete(net, grp);
3099 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3100 tsk->mc_method.rcast = true;
3101 tsk->mc_method.mandatory = true;
3102 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3106 static int tipc_sk_leave(struct tipc_sock *tsk)
3108 struct net *net = sock_net(&tsk->sk);
3109 struct tipc_group *grp = tsk->group;
3110 struct tipc_uaddr ua;
3115 ua.addrtype = TIPC_SERVICE_RANGE;
3116 tipc_group_self(grp, &ua.sr, &scope);
3118 tipc_group_delete(net, grp);
3120 tipc_sk_withdraw(tsk, &ua);
3125 * tipc_setsockopt - set socket option
3126 * @sock: socket structure
3127 * @lvl: option level
3128 * @opt: option identifier
3129 * @ov: pointer to new option value
3130 * @ol: length of option value
3132 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3133 * (to ease compatibility).
3135 * Return: 0 on success, errno otherwise
3137 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3138 sockptr_t ov, unsigned int ol)
3140 struct sock *sk = sock->sk;
3141 struct tipc_sock *tsk = tipc_sk(sk);
3142 struct tipc_group_req mreq;
3146 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3148 if (lvl != SOL_TIPC)
3149 return -ENOPROTOOPT;
3152 case TIPC_IMPORTANCE:
3153 case TIPC_SRC_DROPPABLE:
3154 case TIPC_DEST_DROPPABLE:
3155 case TIPC_CONN_TIMEOUT:
3157 if (ol < sizeof(value))
3159 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3162 case TIPC_GROUP_JOIN:
3163 if (ol < sizeof(mreq))
3165 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3169 if (!sockptr_is_null(ov) || ol)
3176 case TIPC_IMPORTANCE:
3177 res = tsk_set_importance(sk, value);
3179 case TIPC_SRC_DROPPABLE:
3180 if (sock->type != SOCK_STREAM)
3181 tsk_set_unreliable(tsk, value);
3185 case TIPC_DEST_DROPPABLE:
3186 tsk_set_unreturnable(tsk, value);
3188 case TIPC_CONN_TIMEOUT:
3189 tipc_sk(sk)->conn_timeout = value;
3191 case TIPC_MCAST_BROADCAST:
3192 tsk->mc_method.rcast = false;
3193 tsk->mc_method.mandatory = true;
3195 case TIPC_MCAST_REPLICAST:
3196 tsk->mc_method.rcast = true;
3197 tsk->mc_method.mandatory = true;
3199 case TIPC_GROUP_JOIN:
3200 res = tipc_sk_join(tsk, &mreq);
3202 case TIPC_GROUP_LEAVE:
3203 res = tipc_sk_leave(tsk);
3206 tsk->nodelay = !!value;
3219 * tipc_getsockopt - get socket option
3220 * @sock: socket structure
3221 * @lvl: option level
3222 * @opt: option identifier
3223 * @ov: receptacle for option value
3224 * @ol: receptacle for length of option value
3226 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3227 * (to ease compatibility).
3229 * Return: 0 on success, errno otherwise
3231 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3232 char __user *ov, int __user *ol)
3234 struct sock *sk = sock->sk;
3235 struct tipc_sock *tsk = tipc_sk(sk);
3236 struct tipc_service_range seq;
3241 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3242 return put_user(0, ol);
3243 if (lvl != SOL_TIPC)
3244 return -ENOPROTOOPT;
3245 res = get_user(len, ol);
3252 case TIPC_IMPORTANCE:
3253 value = tsk_importance(tsk);
3255 case TIPC_SRC_DROPPABLE:
3256 value = tsk_unreliable(tsk);
3258 case TIPC_DEST_DROPPABLE:
3259 value = tsk_unreturnable(tsk);
3261 case TIPC_CONN_TIMEOUT:
3262 value = tsk->conn_timeout;
3263 /* no need to set "res", since already 0 at this point */
3265 case TIPC_NODE_RECVQ_DEPTH:
3266 value = 0; /* was tipc_queue_size, now obsolete */
3268 case TIPC_SOCK_RECVQ_DEPTH:
3269 value = skb_queue_len(&sk->sk_receive_queue);
3271 case TIPC_SOCK_RECVQ_USED:
3272 value = sk_rmem_alloc_get(sk);
3274 case TIPC_GROUP_JOIN:
3277 tipc_group_self(tsk->group, &seq, &scope);
3287 return res; /* "get" failed */
3289 if (len < sizeof(value))
3292 if (copy_to_user(ov, &value, sizeof(value)))
3295 return put_user(sizeof(value), ol);
3298 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3300 struct net *net = sock_net(sock->sk);
3301 struct tipc_sioc_nodeid_req nr = {0};
3302 struct tipc_sioc_ln_req lnr;
3303 void __user *argp = (void __user *)arg;
3306 case SIOCGETLINKNAME:
3307 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3309 if (!tipc_node_get_linkname(net,
3310 lnr.bearer_id & 0xffff, lnr.peer,
3311 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3312 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3316 return -EADDRNOTAVAIL;
3318 if (copy_from_user(&nr, argp, sizeof(nr)))
3320 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3321 return -EADDRNOTAVAIL;
3322 if (copy_to_user(argp, &nr, sizeof(nr)))
3326 return -ENOIOCTLCMD;
3330 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3332 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3333 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3334 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3336 tsk1->peer.family = AF_TIPC;
3337 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3338 tsk1->peer.scope = TIPC_NODE_SCOPE;
3339 tsk1->peer.addr.id.ref = tsk2->portid;
3340 tsk1->peer.addr.id.node = onode;
3341 tsk2->peer.family = AF_TIPC;
3342 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3343 tsk2->peer.scope = TIPC_NODE_SCOPE;
3344 tsk2->peer.addr.id.ref = tsk1->portid;
3345 tsk2->peer.addr.id.node = onode;
3347 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3348 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3352 /* Protocol switches for the various types of TIPC sockets */
3354 static const struct proto_ops msg_ops = {
3355 .owner = THIS_MODULE,
3357 .release = tipc_release,
3359 .connect = tipc_connect,
3360 .socketpair = tipc_socketpair,
3361 .accept = sock_no_accept,
3362 .getname = tipc_getname,
3364 .ioctl = tipc_ioctl,
3365 .listen = sock_no_listen,
3366 .shutdown = tipc_shutdown,
3367 .setsockopt = tipc_setsockopt,
3368 .getsockopt = tipc_getsockopt,
3369 .sendmsg = tipc_sendmsg,
3370 .recvmsg = tipc_recvmsg,
3371 .mmap = sock_no_mmap,
3372 .sendpage = sock_no_sendpage
3375 static const struct proto_ops packet_ops = {
3376 .owner = THIS_MODULE,
3378 .release = tipc_release,
3380 .connect = tipc_connect,
3381 .socketpair = tipc_socketpair,
3382 .accept = tipc_accept,
3383 .getname = tipc_getname,
3385 .ioctl = tipc_ioctl,
3386 .listen = tipc_listen,
3387 .shutdown = tipc_shutdown,
3388 .setsockopt = tipc_setsockopt,
3389 .getsockopt = tipc_getsockopt,
3390 .sendmsg = tipc_send_packet,
3391 .recvmsg = tipc_recvmsg,
3392 .mmap = sock_no_mmap,
3393 .sendpage = sock_no_sendpage
3396 static const struct proto_ops stream_ops = {
3397 .owner = THIS_MODULE,
3399 .release = tipc_release,
3401 .connect = tipc_connect,
3402 .socketpair = tipc_socketpair,
3403 .accept = tipc_accept,
3404 .getname = tipc_getname,
3406 .ioctl = tipc_ioctl,
3407 .listen = tipc_listen,
3408 .shutdown = tipc_shutdown,
3409 .setsockopt = tipc_setsockopt,
3410 .getsockopt = tipc_getsockopt,
3411 .sendmsg = tipc_sendstream,
3412 .recvmsg = tipc_recvstream,
3413 .mmap = sock_no_mmap,
3414 .sendpage = sock_no_sendpage
3417 static const struct net_proto_family tipc_family_ops = {
3418 .owner = THIS_MODULE,
3420 .create = tipc_sk_create
3423 static struct proto tipc_proto = {
3425 .owner = THIS_MODULE,
3426 .obj_size = sizeof(struct tipc_sock),
3427 .sysctl_rmem = sysctl_tipc_rmem
3431 * tipc_socket_init - initialize TIPC socket interface
3433 * Return: 0 on success, errno otherwise
3435 int tipc_socket_init(void)
3439 res = proto_register(&tipc_proto, 1);
3441 pr_err("Failed to register TIPC protocol type\n");
3445 res = sock_register(&tipc_family_ops);
3447 pr_err("Failed to register TIPC socket type\n");
3448 proto_unregister(&tipc_proto);
3456 * tipc_socket_stop - stop TIPC socket interface
3458 void tipc_socket_stop(void)
3460 sock_unregister(tipc_family_ops.family);
3461 proto_unregister(&tipc_proto);
3464 /* Caller should hold socket lock for the passed tipc socket. */
3465 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3467 u32 peer_node, peer_port;
3468 u32 conn_type, conn_instance;
3469 struct nlattr *nest;
3471 peer_node = tsk_peer_node(tsk);
3472 peer_port = tsk_peer_port(tsk);
3473 conn_type = msg_nametype(&tsk->phdr);
3474 conn_instance = msg_nameinst(&tsk->phdr);
3475 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3479 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3481 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3484 if (tsk->conn_addrtype != 0) {
3485 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3487 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3489 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3492 nla_nest_end(skb, nest);
3497 nla_nest_cancel(skb, nest);
3502 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3505 struct net *net = sock_net(skb->sk);
3506 struct sock *sk = &tsk->sk;
3508 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3509 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3512 if (tipc_sk_connected(sk)) {
3513 if (__tipc_nl_add_sk_con(skb, tsk))
3515 } else if (!list_empty(&tsk->publications)) {
3516 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3522 /* Caller should hold socket lock for the passed tipc socket. */
3523 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3524 struct tipc_sock *tsk)
3526 struct nlattr *attrs;
3529 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3530 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3534 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3536 goto genlmsg_cancel;
3538 if (__tipc_nl_add_sk_info(skb, tsk))
3539 goto attr_msg_cancel;
3541 nla_nest_end(skb, attrs);
3542 genlmsg_end(skb, hdr);
3547 nla_nest_cancel(skb, attrs);
3549 genlmsg_cancel(skb, hdr);
3554 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3555 int (*skb_handler)(struct sk_buff *skb,
3556 struct netlink_callback *cb,
3557 struct tipc_sock *tsk))
3559 struct rhashtable_iter *iter = (void *)cb->args[4];
3560 struct tipc_sock *tsk;
3563 rhashtable_walk_start(iter);
3564 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3567 if (err == -EAGAIN) {
3574 sock_hold(&tsk->sk);
3575 rhashtable_walk_stop(iter);
3576 lock_sock(&tsk->sk);
3577 err = skb_handler(skb, cb, tsk);
3579 release_sock(&tsk->sk);
3583 release_sock(&tsk->sk);
3584 rhashtable_walk_start(iter);
3587 rhashtable_walk_stop(iter);
3591 EXPORT_SYMBOL(tipc_nl_sk_walk);
3593 int tipc_dump_start(struct netlink_callback *cb)
3595 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3597 EXPORT_SYMBOL(tipc_dump_start);
3599 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3601 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3602 struct rhashtable_iter *iter = (void *)cb->args[4];
3603 struct tipc_net *tn = tipc_net(net);
3606 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3610 cb->args[4] = (long)iter;
3613 rhashtable_walk_enter(&tn->sk_rht, iter);
3617 int tipc_dump_done(struct netlink_callback *cb)
3619 struct rhashtable_iter *hti = (void *)cb->args[4];
3621 rhashtable_walk_exit(hti);
3625 EXPORT_SYMBOL(tipc_dump_done);
3627 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3628 struct tipc_sock *tsk, u32 sk_filter_state,
3629 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3631 struct sock *sk = &tsk->sk;
3632 struct nlattr *attrs;
3633 struct nlattr *stat;
3635 /*filter response w.r.t sk_state*/
3636 if (!(sk_filter_state & (1 << sk->sk_state)))
3639 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3643 if (__tipc_nl_add_sk_info(skb, tsk))
3644 goto attr_msg_cancel;
3646 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3647 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3648 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3649 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3650 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3652 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3653 tipc_diag_gen_cookie(sk),
3655 goto attr_msg_cancel;
3657 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3659 goto attr_msg_cancel;
3661 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3662 skb_queue_len(&sk->sk_receive_queue)) ||
3663 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3664 skb_queue_len(&sk->sk_write_queue)) ||
3665 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3666 atomic_read(&sk->sk_drops)))
3667 goto stat_msg_cancel;
3669 if (tsk->cong_link_cnt &&
3670 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3671 goto stat_msg_cancel;
3673 if (tsk_conn_cong(tsk) &&
3674 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3675 goto stat_msg_cancel;
3677 nla_nest_end(skb, stat);
3680 if (tipc_group_fill_sock_diag(tsk->group, skb))
3681 goto stat_msg_cancel;
3683 nla_nest_end(skb, attrs);
3688 nla_nest_cancel(skb, stat);
3690 nla_nest_cancel(skb, attrs);
3694 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3696 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3698 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3701 /* Caller should hold socket lock for the passed tipc socket. */
3702 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3703 struct netlink_callback *cb,
3704 struct publication *publ)
3707 struct nlattr *attrs;
3709 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3710 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3714 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3716 goto genlmsg_cancel;
3718 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3719 goto attr_msg_cancel;
3720 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3721 goto attr_msg_cancel;
3722 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3723 goto attr_msg_cancel;
3724 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3725 goto attr_msg_cancel;
3727 nla_nest_end(skb, attrs);
3728 genlmsg_end(skb, hdr);
3733 nla_nest_cancel(skb, attrs);
3735 genlmsg_cancel(skb, hdr);
3740 /* Caller should hold socket lock for the passed tipc socket. */
3741 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3742 struct netlink_callback *cb,
3743 struct tipc_sock *tsk, u32 *last_publ)
3746 struct publication *p;
3749 list_for_each_entry(p, &tsk->publications, binding_sock) {
3750 if (p->key == *last_publ)
3753 if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3754 /* We never set seq or call nl_dump_check_consistent()
3755 * this means that setting prev_seq here will cause the
3756 * consistence check to fail in the netlink callback
3757 * handler. Resulting in the last NLMSG_DONE message
3758 * having the NLM_F_DUMP_INTR flag set.
3765 p = list_first_entry(&tsk->publications, struct publication,
3769 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3770 err = __tipc_nl_add_sk_publ(skb, cb, p);
3772 *last_publ = p->key;
3781 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3784 u32 tsk_portid = cb->args[0];
3785 u32 last_publ = cb->args[1];
3786 u32 done = cb->args[2];
3787 struct net *net = sock_net(skb->sk);
3788 struct tipc_sock *tsk;
3791 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3792 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3794 if (!attrs[TIPC_NLA_SOCK])
3797 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3798 attrs[TIPC_NLA_SOCK],
3799 tipc_nl_sock_policy, NULL);
3803 if (!sock[TIPC_NLA_SOCK_REF])
3806 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3812 tsk = tipc_sk_lookup(net, tsk_portid);
3816 lock_sock(&tsk->sk);
3817 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3820 release_sock(&tsk->sk);
3823 cb->args[0] = tsk_portid;
3824 cb->args[1] = last_publ;
3831 * tipc_sk_filtering - check if a socket should be traced
3832 * @sk: the socket to be examined
3834 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3835 * (portid, sock type, name type, name lower, name upper)
3837 * Return: true if the socket meets the socket tuple data
3838 * (value 0 = 'any') or when there is no tuple set (all = 0),
3841 bool tipc_sk_filtering(struct sock *sk)
3843 struct tipc_sock *tsk;
3844 struct publication *p;
3845 u32 _port, _sktype, _type, _lower, _upper;
3846 u32 type = 0, lower = 0, upper = 0;
3853 _port = sysctl_tipc_sk_filter[0];
3854 _sktype = sysctl_tipc_sk_filter[1];
3855 _type = sysctl_tipc_sk_filter[2];
3856 _lower = sysctl_tipc_sk_filter[3];
3857 _upper = sysctl_tipc_sk_filter[4];
3859 if (!_port && !_sktype && !_type && !_lower && !_upper)
3863 return (_port == tsk->portid);
3865 if (_sktype && _sktype != sk->sk_type)
3868 if (tsk->published) {
3869 p = list_first_entry_or_null(&tsk->publications,
3870 struct publication, binding_sock);
3873 lower = p->sr.lower;
3874 upper = p->sr.upper;
3878 if (!tipc_sk_type_connectionless(sk)) {
3879 type = msg_nametype(&tsk->phdr);
3880 lower = msg_nameinst(&tsk->phdr);
3884 if ((_type && _type != type) || (_lower && _lower != lower) ||
3885 (_upper && _upper != upper))
3891 u32 tipc_sock_get_portid(struct sock *sk)
3893 return (sk) ? (tipc_sk(sk))->portid : 0;
3897 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3898 * both the rcv and backlog queues are considered
3899 * @sk: tipc sk to be checked
3900 * @skb: tipc msg to be checked
3902 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3905 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3907 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3908 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3909 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3911 return (qsize > lim * 90 / 100);
3915 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3916 * only the rcv queue is considered
3917 * @sk: tipc sk to be checked
3918 * @skb: tipc msg to be checked
3920 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3923 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3925 unsigned int lim = rcvbuf_limit(sk, skb);
3926 unsigned int qsize = sk_rmem_alloc_get(sk);
3928 return (qsize > lim * 90 / 100);
3932 * tipc_sk_dump - dump TIPC socket
3933 * @sk: tipc sk to be dumped
3934 * @dqueues: bitmask to decide if any socket queue to be dumped?
3935 * - TIPC_DUMP_NONE: don't dump socket queues
3936 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3937 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3938 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3939 * - TIPC_DUMP_ALL: dump all the socket queues above
3940 * @buf: returned buffer of dump data in format
3942 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3945 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3946 u32 conn_type, conn_instance;
3947 struct tipc_sock *tsk;
3948 struct publication *p;
3952 i += scnprintf(buf, sz, "sk data: (null)\n");
3957 tsk_connected = !tipc_sk_type_connectionless(sk);
3959 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3960 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3961 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3962 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3963 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3964 if (tsk_connected) {
3965 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3966 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3967 conn_type = msg_nametype(&tsk->phdr);
3968 conn_instance = msg_nameinst(&tsk->phdr);
3969 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3970 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3972 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3973 if (tsk->published) {
3974 p = list_first_entry_or_null(&tsk->publications,
3975 struct publication, binding_sock);
3976 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3977 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3978 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3980 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3981 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3982 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3983 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3984 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3985 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3986 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3987 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3988 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3989 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3990 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3991 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3992 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3993 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3995 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3996 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3997 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
4000 if (dqueues & TIPC_DUMP_SK_RCVQ) {
4001 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
4002 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4005 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4006 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4007 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4008 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4009 i += scnprintf(buf + i, sz - i, " tail ");
4010 i += tipc_skb_dump(sk->sk_backlog.tail, false,