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)) {
506 pr_warn("Socket create failed; port number exhausted\n");
510 /* Ensure tsk is visible before we read own_addr. */
513 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
514 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
516 msg_set_origport(msg, tsk->portid);
517 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
519 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
520 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
521 sk->sk_data_ready = tipc_data_ready;
522 sk->sk_write_space = tipc_write_space;
523 sk->sk_destruct = tipc_sock_destruct;
524 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
525 tsk->group_is_open = true;
526 atomic_set(&tsk->dupl_rcvcnt, 0);
528 /* Start out with safe limits until we receive an advertised window */
529 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
530 tsk->rcv_win = tsk->snd_win;
532 if (tipc_sk_type_connectionless(sk)) {
533 tsk_set_unreturnable(tsk, true);
534 if (sock->type == SOCK_DGRAM)
535 tsk_set_unreliable(tsk, true);
537 __skb_queue_head_init(&tsk->mc_method.deferredq);
538 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
542 static void tipc_sk_callback(struct rcu_head *head)
544 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
549 /* Caller should hold socket lock for the socket. */
550 static void __tipc_shutdown(struct socket *sock, int error)
552 struct sock *sk = sock->sk;
553 struct tipc_sock *tsk = tipc_sk(sk);
554 struct net *net = sock_net(sk);
555 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
556 u32 dnode = tsk_peer_node(tsk);
559 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
560 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
561 !tsk_conn_cong(tsk)));
563 /* Push out delayed messages if in Nagle mode */
564 tipc_sk_push_backlog(tsk, false);
565 /* Remove pending SYN */
566 __skb_queue_purge(&sk->sk_write_queue);
568 /* Remove partially received buffer if any */
569 skb = skb_peek(&sk->sk_receive_queue);
570 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
571 __skb_unlink(skb, &sk->sk_receive_queue);
575 /* Reject all unreceived messages if connectionless */
576 if (tipc_sk_type_connectionless(sk)) {
577 tsk_rej_rx_queue(sk, error);
581 switch (sk->sk_state) {
582 case TIPC_CONNECTING:
583 case TIPC_ESTABLISHED:
584 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
585 tipc_node_remove_conn(net, dnode, tsk->portid);
586 /* Send a FIN+/- to its peer */
587 skb = __skb_dequeue(&sk->sk_receive_queue);
589 __skb_queue_purge(&sk->sk_receive_queue);
590 tipc_sk_respond(sk, skb, error);
593 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
594 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
595 tsk_own_node(tsk), tsk_peer_port(tsk),
598 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
601 /* Reject all SYN messages */
602 tsk_rej_rx_queue(sk, error);
605 __skb_queue_purge(&sk->sk_receive_queue);
611 * tipc_release - destroy a TIPC socket
612 * @sock: socket to destroy
614 * This routine cleans up any messages that are still queued on the socket.
615 * For DGRAM and RDM socket types, all queued messages are rejected.
616 * For SEQPACKET and STREAM socket types, the first message is rejected
617 * and any others are discarded. (If the first message on a STREAM socket
618 * is partially-read, it is discarded and the next one is rejected instead.)
620 * NOTE: Rejected messages are not necessarily returned to the sender! They
621 * are returned or discarded according to the "destination droppable" setting
622 * specified for the message by the sender.
624 * Return: 0 on success, errno otherwise
626 static int tipc_release(struct socket *sock)
628 struct sock *sk = sock->sk;
629 struct tipc_sock *tsk;
632 * Exit if socket isn't fully initialized (occurs when a failed accept()
633 * releases a pre-allocated child socket that was never used)
641 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
642 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
643 sk->sk_shutdown = SHUTDOWN_MASK;
645 tipc_sk_withdraw(tsk, NULL);
646 __skb_queue_purge(&tsk->mc_method.deferredq);
647 sk_stop_timer(sk, &sk->sk_timer);
651 /* Reject any messages that accumulated in backlog queue */
653 tipc_dest_list_purge(&tsk->cong_links);
654 tsk->cong_link_cnt = 0;
655 call_rcu(&tsk->rcu, tipc_sk_callback);
662 * __tipc_bind - associate or disassocate TIPC name(s) with a socket
663 * @sock: socket structure
664 * @skaddr: socket address describing name(s) and desired operation
665 * @alen: size of socket address data structure
667 * Name and name sequence binding are indicated using a positive scope value;
668 * a negative scope value unbinds the specified name. Specifying no name
669 * (i.e. a socket address length of 0) unbinds all names from the socket.
671 * Return: 0 on success, errno otherwise
673 * NOTE: This routine doesn't need to take the socket lock since it doesn't
674 * access any non-constant socket information.
676 static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
678 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
679 struct tipc_sock *tsk = tipc_sk(sock->sk);
683 return tipc_sk_withdraw(tsk, NULL);
685 if (ua->addrtype == TIPC_SERVICE_ADDR) {
686 ua->addrtype = TIPC_SERVICE_RANGE;
687 ua->sr.upper = ua->sr.lower;
691 ua->scope = -ua->scope;
693 /* Users may still use deprecated TIPC_ZONE_SCOPE */
694 if (ua->scope != TIPC_NODE_SCOPE)
695 ua->scope = TIPC_CLUSTER_SCOPE;
701 return tipc_sk_withdraw(tsk, ua);
702 return tipc_sk_publish(tsk, ua);
705 int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
710 res = __tipc_bind(sock, skaddr, alen);
711 release_sock(sock->sk);
715 static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
717 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
718 u32 atype = ua->addrtype;
721 if (!tipc_uaddr_valid(ua, alen))
723 if (atype == TIPC_SOCKET_ADDR)
724 return -EAFNOSUPPORT;
725 if (ua->sr.type < TIPC_RESERVED_TYPES) {
726 pr_warn_once("Can't bind to reserved service type %u\n",
731 return tipc_sk_bind(sock, skaddr, alen);
735 * tipc_getname - get port ID of socket or peer socket
736 * @sock: socket structure
737 * @uaddr: area for returned socket address
738 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
740 * Return: 0 on success, errno otherwise
742 * NOTE: This routine doesn't need to take the socket lock since it only
743 * accesses socket information that is unchanging (or which changes in
744 * a completely predictable manner).
746 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
749 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
750 struct sock *sk = sock->sk;
751 struct tipc_sock *tsk = tipc_sk(sk);
753 memset(addr, 0, sizeof(*addr));
755 if ((!tipc_sk_connected(sk)) &&
756 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
758 addr->addr.id.ref = tsk_peer_port(tsk);
759 addr->addr.id.node = tsk_peer_node(tsk);
761 addr->addr.id.ref = tsk->portid;
762 addr->addr.id.node = tipc_own_addr(sock_net(sk));
765 addr->addrtype = TIPC_SOCKET_ADDR;
766 addr->family = AF_TIPC;
768 addr->addr.name.domain = 0;
770 return sizeof(*addr);
774 * tipc_poll - read and possibly block on pollmask
775 * @file: file structure associated with the socket
776 * @sock: socket for which to calculate the poll bits
779 * Return: pollmask value
782 * It appears that the usual socket locking mechanisms are not useful here
783 * since the pollmask info is potentially out-of-date the moment this routine
784 * exits. TCP and other protocols seem to rely on higher level poll routines
785 * to handle any preventable race conditions, so TIPC will do the same ...
787 * IMPORTANT: The fact that a read or write operation is indicated does NOT
788 * imply that the operation will succeed, merely that it should be performed
789 * and will not block.
791 static __poll_t tipc_poll(struct file *file, struct socket *sock,
794 struct sock *sk = sock->sk;
795 struct tipc_sock *tsk = tipc_sk(sk);
796 __poll_t revents = 0;
798 sock_poll_wait(file, sock, wait);
799 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
801 if (sk->sk_shutdown & RCV_SHUTDOWN)
802 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
803 if (sk->sk_shutdown == SHUTDOWN_MASK)
806 switch (sk->sk_state) {
807 case TIPC_ESTABLISHED:
808 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
812 case TIPC_CONNECTING:
813 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
814 revents |= EPOLLIN | EPOLLRDNORM;
817 if (tsk->group_is_open && !tsk->cong_link_cnt)
819 if (!tipc_sk_type_connectionless(sk))
821 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
823 revents |= EPOLLIN | EPOLLRDNORM;
825 case TIPC_DISCONNECTING:
826 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
833 * tipc_sendmcast - send multicast message
834 * @sock: socket structure
835 * @ua: destination address struct
836 * @msg: message to send
837 * @dlen: length of data to send
838 * @timeout: timeout to wait for wakeup
840 * Called from function tipc_sendmsg(), which has done all sanity checks
841 * Return: the number of bytes sent on success, or errno
843 static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
844 struct msghdr *msg, size_t dlen, long timeout)
846 struct sock *sk = sock->sk;
847 struct tipc_sock *tsk = tipc_sk(sk);
848 struct tipc_msg *hdr = &tsk->phdr;
849 struct net *net = sock_net(sk);
850 int mtu = tipc_bcast_get_mtu(net);
851 struct sk_buff_head pkts;
852 struct tipc_nlist dsts;
858 /* Block or return if any destination link is congested */
859 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
863 /* Lookup destination nodes */
864 tipc_nlist_init(&dsts, tipc_own_addr(net));
865 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
866 if (!dsts.local && !dsts.remote)
867 return -EHOSTUNREACH;
869 /* Build message header */
870 msg_set_type(hdr, TIPC_MCAST_MSG);
871 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
872 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
873 msg_set_destport(hdr, 0);
874 msg_set_destnode(hdr, 0);
875 msg_set_nametype(hdr, ua->sr.type);
876 msg_set_namelower(hdr, ua->sr.lower);
877 msg_set_nameupper(hdr, ua->sr.upper);
879 /* Build message as chain of buffers */
880 __skb_queue_head_init(&pkts);
881 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
883 /* Send message if build was successful */
884 if (unlikely(rc == dlen)) {
885 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
886 TIPC_DUMP_SK_SNDQ, " ");
887 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
888 &tsk->cong_link_cnt);
891 tipc_nlist_purge(&dsts);
893 return rc ? rc : dlen;
897 * tipc_send_group_msg - send a message to a member in the group
898 * @net: network namespace
900 * @m: message to send
902 * @dnode: destination node
903 * @dport: destination port
904 * @dlen: total length of message data
906 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
907 struct msghdr *m, struct tipc_member *mb,
908 u32 dnode, u32 dport, int dlen)
910 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
911 struct tipc_mc_method *method = &tsk->mc_method;
912 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
913 struct tipc_msg *hdr = &tsk->phdr;
914 struct sk_buff_head pkts;
917 /* Complete message header */
918 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
919 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
920 msg_set_destport(hdr, dport);
921 msg_set_destnode(hdr, dnode);
922 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
924 /* Build message as chain of buffers */
925 __skb_queue_head_init(&pkts);
926 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
927 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
928 if (unlikely(rc != dlen))
932 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
933 if (unlikely(rc == -ELINKCONG)) {
934 tipc_dest_push(&tsk->cong_links, dnode, 0);
935 tsk->cong_link_cnt++;
938 /* Update send window */
939 tipc_group_update_member(mb, blks);
941 /* A broadcast sent within next EXPIRE period must follow same path */
942 method->rcast = true;
943 method->mandatory = true;
948 * tipc_send_group_unicast - send message to a member in the group
949 * @sock: socket structure
950 * @m: message to send
951 * @dlen: total length of message data
952 * @timeout: timeout to wait for wakeup
954 * Called from function tipc_sendmsg(), which has done all sanity checks
955 * Return: the number of bytes sent on success, or errno
957 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
958 int dlen, long timeout)
960 struct sock *sk = sock->sk;
961 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
962 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
963 struct tipc_sock *tsk = tipc_sk(sk);
964 struct net *net = sock_net(sk);
965 struct tipc_member *mb = NULL;
972 return -EHOSTUNREACH;
974 /* Block or return if destination link or member is congested */
975 rc = tipc_wait_for_cond(sock, &timeout,
976 !tipc_dest_find(&tsk->cong_links, node, 0) &&
978 !tipc_group_cong(tsk->group, node, port, blks,
984 return -EHOSTUNREACH;
986 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
988 return rc ? rc : dlen;
992 * tipc_send_group_anycast - send message to any member with given identity
993 * @sock: socket structure
994 * @m: message to send
995 * @dlen: total length of message data
996 * @timeout: timeout to wait for wakeup
998 * Called from function tipc_sendmsg(), which has done all sanity checks
999 * Return: the number of bytes sent on success, or errno
1001 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1002 int dlen, long timeout)
1004 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1005 struct sock *sk = sock->sk;
1006 struct tipc_sock *tsk = tipc_sk(sk);
1007 struct list_head *cong_links = &tsk->cong_links;
1008 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1009 struct tipc_msg *hdr = &tsk->phdr;
1010 struct tipc_member *first = NULL;
1011 struct tipc_member *mbr = NULL;
1012 struct net *net = sock_net(sk);
1013 u32 node, port, exclude;
1014 struct list_head dsts;
1019 INIT_LIST_HEAD(&dsts);
1020 ua->sa.type = msg_nametype(hdr);
1021 ua->scope = msg_lookup_scope(hdr);
1023 while (++lookups < 4) {
1024 exclude = tipc_group_exclude(tsk->group);
1028 /* Look for a non-congested destination member, if any */
1030 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1032 return -EHOSTUNREACH;
1033 tipc_dest_pop(&dsts, &node, &port);
1034 cong = tipc_group_cong(tsk->group, node, port, blks,
1044 /* Start over if destination was not in member list */
1048 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1051 /* Block or return if destination link or member is congested */
1052 rc = tipc_wait_for_cond(sock, &timeout,
1053 !tipc_dest_find(cong_links, node, 0) &&
1055 !tipc_group_cong(tsk->group, node, port,
1060 /* Send, unless destination disappeared while waiting */
1065 if (unlikely(lookups >= 4))
1066 return -EHOSTUNREACH;
1068 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1070 return rc ? rc : dlen;
1074 * tipc_send_group_bcast - send message to all members in communication group
1075 * @sock: socket structure
1076 * @m: message to send
1077 * @dlen: total length of message data
1078 * @timeout: timeout to wait for wakeup
1080 * Called from function tipc_sendmsg(), which has done all sanity checks
1081 * Return: the number of bytes sent on success, or errno
1083 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1084 int dlen, long timeout)
1086 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1087 struct sock *sk = sock->sk;
1088 struct net *net = sock_net(sk);
1089 struct tipc_sock *tsk = tipc_sk(sk);
1090 struct tipc_nlist *dsts;
1091 struct tipc_mc_method *method = &tsk->mc_method;
1092 bool ack = method->mandatory && method->rcast;
1093 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1094 struct tipc_msg *hdr = &tsk->phdr;
1095 int mtu = tipc_bcast_get_mtu(net);
1096 struct sk_buff_head pkts;
1097 int rc = -EHOSTUNREACH;
1099 /* Block or return if any destination link or member is congested */
1100 rc = tipc_wait_for_cond(sock, &timeout,
1101 !tsk->cong_link_cnt && tsk->group &&
1102 !tipc_group_bc_cong(tsk->group, blks));
1106 dsts = tipc_group_dests(tsk->group);
1107 if (!dsts->local && !dsts->remote)
1108 return -EHOSTUNREACH;
1110 /* Complete message header */
1112 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1113 msg_set_nameinst(hdr, ua->sa.instance);
1115 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1116 msg_set_nameinst(hdr, 0);
1118 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1119 msg_set_destport(hdr, 0);
1120 msg_set_destnode(hdr, 0);
1121 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1123 /* Avoid getting stuck with repeated forced replicasts */
1124 msg_set_grp_bc_ack_req(hdr, ack);
1126 /* Build message as chain of buffers */
1127 __skb_queue_head_init(&pkts);
1128 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1129 if (unlikely(rc != dlen))
1133 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1137 /* Update broadcast sequence number and send windows */
1138 tipc_group_update_bc_members(tsk->group, blks, ack);
1140 /* Broadcast link is now free to choose method for next broadcast */
1141 method->mandatory = false;
1142 method->expires = jiffies;
1148 * tipc_send_group_mcast - send message to all members with given identity
1149 * @sock: socket structure
1150 * @m: message to send
1151 * @dlen: total length of message data
1152 * @timeout: timeout to wait for wakeup
1154 * Called from function tipc_sendmsg(), which has done all sanity checks
1155 * Return: the number of bytes sent on success, or errno
1157 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1158 int dlen, long timeout)
1160 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1161 struct sock *sk = sock->sk;
1162 struct tipc_sock *tsk = tipc_sk(sk);
1163 struct tipc_group *grp = tsk->group;
1164 struct tipc_msg *hdr = &tsk->phdr;
1165 struct net *net = sock_net(sk);
1166 struct list_head dsts;
1167 u32 dstcnt, exclude;
1169 INIT_LIST_HEAD(&dsts);
1170 ua->sa.type = msg_nametype(hdr);
1171 ua->scope = msg_lookup_scope(hdr);
1172 exclude = tipc_group_exclude(grp);
1174 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1175 return -EHOSTUNREACH;
1178 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1179 return tipc_send_group_unicast(sock, m, dlen, timeout);
1182 tipc_dest_list_purge(&dsts);
1183 return tipc_send_group_bcast(sock, m, dlen, timeout);
1187 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1188 * @net: the associated network namespace
1189 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1190 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1192 * Multi-threaded: parallel calls with reference to same queues may occur
1194 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1195 struct sk_buff_head *inputq)
1197 u32 self = tipc_own_addr(net);
1198 struct sk_buff *skb, *_skb;
1200 struct sk_buff_head tmpq;
1201 struct list_head dports;
1202 struct tipc_msg *hdr;
1203 struct tipc_uaddr ua;
1204 int user, mtyp, hlen;
1206 __skb_queue_head_init(&tmpq);
1207 INIT_LIST_HEAD(&dports);
1208 ua.addrtype = TIPC_SERVICE_RANGE;
1210 /* tipc_skb_peek() increments the head skb's reference counter */
1211 skb = tipc_skb_peek(arrvq, &inputq->lock);
1212 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1214 user = msg_user(hdr);
1215 mtyp = msg_type(hdr);
1216 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1217 onode = msg_orignode(hdr);
1218 ua.sr.type = msg_nametype(hdr);
1219 ua.sr.lower = msg_namelower(hdr);
1220 ua.sr.upper = msg_nameupper(hdr);
1222 ua.scope = TIPC_ANY_SCOPE;
1224 ua.scope = TIPC_CLUSTER_SCOPE;
1226 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1227 spin_lock_bh(&inputq->lock);
1228 if (skb_peek(arrvq) == skb) {
1229 __skb_dequeue(arrvq);
1230 __skb_queue_tail(inputq, skb);
1233 spin_unlock_bh(&inputq->lock);
1237 /* Group messages require exact scope match */
1238 if (msg_in_group(hdr)) {
1241 ua.scope = msg_lookup_scope(hdr);
1244 /* Create destination port list: */
1245 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1247 /* Clone message per destination */
1248 while (tipc_dest_pop(&dports, NULL, &portid)) {
1249 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1251 msg_set_destport(buf_msg(_skb), portid);
1252 __skb_queue_tail(&tmpq, _skb);
1255 pr_warn("Failed to clone mcast rcv buffer\n");
1257 /* Append clones to inputq only if skb is still head of arrvq */
1258 spin_lock_bh(&inputq->lock);
1259 if (skb_peek(arrvq) == skb) {
1260 skb_queue_splice_tail_init(&tmpq, inputq);
1261 /* Decrement the skb's refcnt */
1262 kfree_skb(__skb_dequeue(arrvq));
1264 spin_unlock_bh(&inputq->lock);
1265 __skb_queue_purge(&tmpq);
1268 tipc_sk_rcv(net, inputq);
1271 /* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1272 * when socket is in Nagle mode
1274 static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1276 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1277 struct sk_buff *skb = skb_peek_tail(txq);
1278 struct net *net = sock_net(&tsk->sk);
1279 u32 dnode = tsk_peer_node(tsk);
1283 tsk->pkt_cnt += skb_queue_len(txq);
1284 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1286 if (tsk->nagle_start < NAGLE_START_MAX)
1287 tsk->nagle_start *= 2;
1288 tsk->expect_ack = false;
1289 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1290 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1293 tsk->nagle_start = NAGLE_START_INIT;
1295 msg_set_ack_required(buf_msg(skb));
1296 tsk->expect_ack = true;
1298 tsk->expect_ack = false;
1305 if (!skb || tsk->cong_link_cnt)
1308 /* Do not send SYN again after congestion */
1309 if (msg_is_syn(buf_msg(skb)))
1313 tsk->pkt_cnt += skb_queue_len(txq);
1314 tsk->snt_unacked += tsk->snd_backlog;
1315 tsk->snd_backlog = 0;
1316 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1317 if (rc == -ELINKCONG)
1318 tsk->cong_link_cnt = 1;
1322 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1323 * @tsk: receiving socket
1324 * @skb: pointer to message buffer.
1325 * @inputq: buffer list containing the buffers
1326 * @xmitq: output message area
1328 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1329 struct sk_buff_head *inputq,
1330 struct sk_buff_head *xmitq)
1332 struct tipc_msg *hdr = buf_msg(skb);
1333 u32 onode = tsk_own_node(tsk);
1334 struct sock *sk = &tsk->sk;
1335 int mtyp = msg_type(hdr);
1338 /* Ignore if connection cannot be validated: */
1339 if (!tsk_peer_msg(tsk, hdr)) {
1340 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1344 if (unlikely(msg_errcode(hdr))) {
1345 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1346 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1347 tsk_peer_port(tsk));
1348 sk->sk_state_change(sk);
1350 /* State change is ignored if socket already awake,
1351 * - convert msg to abort msg and add to inqueue
1353 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1354 msg_set_type(hdr, TIPC_CONN_MSG);
1355 msg_set_size(hdr, BASIC_H_SIZE);
1356 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1357 __skb_queue_tail(inputq, skb);
1361 tsk->probe_unacked = false;
1363 if (mtyp == CONN_PROBE) {
1364 msg_set_type(hdr, CONN_PROBE_REPLY);
1365 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1366 __skb_queue_tail(xmitq, skb);
1368 } else if (mtyp == CONN_ACK) {
1369 was_cong = tsk_conn_cong(tsk);
1370 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1371 tsk->snt_unacked -= msg_conn_ack(hdr);
1372 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1373 tsk->snd_win = msg_adv_win(hdr);
1374 if (was_cong && !tsk_conn_cong(tsk))
1375 sk->sk_write_space(sk);
1376 } else if (mtyp != CONN_PROBE_REPLY) {
1377 pr_warn("Received unknown CONN_PROTO msg\n");
1384 * tipc_sendmsg - send message in connectionless manner
1385 * @sock: socket structure
1386 * @m: message to send
1387 * @dsz: amount of user data to be sent
1389 * Message must have an destination specified explicitly.
1390 * Used for SOCK_RDM and SOCK_DGRAM messages,
1391 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1392 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1394 * Return: the number of bytes sent on success, or errno otherwise
1396 static int tipc_sendmsg(struct socket *sock,
1397 struct msghdr *m, size_t dsz)
1399 struct sock *sk = sock->sk;
1403 ret = __tipc_sendmsg(sock, m, dsz);
1409 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1411 struct sock *sk = sock->sk;
1412 struct net *net = sock_net(sk);
1413 struct tipc_sock *tsk = tipc_sk(sk);
1414 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1415 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1416 struct list_head *clinks = &tsk->cong_links;
1417 bool syn = !tipc_sk_type_connectionless(sk);
1418 struct tipc_group *grp = tsk->group;
1419 struct tipc_msg *hdr = &tsk->phdr;
1420 struct tipc_socket_addr skaddr;
1421 struct sk_buff_head pkts;
1424 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1428 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1430 atype = ua->addrtype;
1433 /* If socket belongs to a communication group follow other paths */
1436 return tipc_send_group_bcast(sock, m, dlen, timeout);
1437 if (atype == TIPC_SERVICE_ADDR)
1438 return tipc_send_group_anycast(sock, m, dlen, timeout);
1439 if (atype == TIPC_SOCKET_ADDR)
1440 return tipc_send_group_unicast(sock, m, dlen, timeout);
1441 if (atype == TIPC_SERVICE_RANGE)
1442 return tipc_send_group_mcast(sock, m, dlen, timeout);
1447 ua = (struct tipc_uaddr *)&tsk->peer;
1448 if (!syn && ua->family != AF_TIPC)
1449 return -EDESTADDRREQ;
1450 atype = ua->addrtype;
1453 if (unlikely(syn)) {
1454 if (sk->sk_state == TIPC_LISTEN)
1456 if (sk->sk_state != TIPC_OPEN)
1460 if (atype == TIPC_SERVICE_ADDR)
1461 tsk->conn_addrtype = atype;
1462 msg_set_syn(hdr, 1);
1465 memset(&skaddr, 0, sizeof(skaddr));
1467 /* Determine destination */
1468 if (atype == TIPC_SERVICE_RANGE) {
1469 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1470 } else if (atype == TIPC_SERVICE_ADDR) {
1471 skaddr.node = ua->lookup_node;
1472 ua->scope = tipc_node2scope(skaddr.node);
1473 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1474 return -EHOSTUNREACH;
1475 } else if (atype == TIPC_SOCKET_ADDR) {
1481 /* Block or return if destination link is congested */
1482 rc = tipc_wait_for_cond(sock, &timeout,
1483 !tipc_dest_find(clinks, skaddr.node, 0));
1487 /* Finally build message header */
1488 msg_set_destnode(hdr, skaddr.node);
1489 msg_set_destport(hdr, skaddr.ref);
1490 if (atype == TIPC_SERVICE_ADDR) {
1491 msg_set_type(hdr, TIPC_NAMED_MSG);
1492 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1493 msg_set_nametype(hdr, ua->sa.type);
1494 msg_set_nameinst(hdr, ua->sa.instance);
1495 msg_set_lookup_scope(hdr, ua->scope);
1496 } else { /* TIPC_SOCKET_ADDR */
1497 msg_set_type(hdr, TIPC_DIRECT_MSG);
1498 msg_set_lookup_scope(hdr, 0);
1499 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1502 /* Add message body */
1503 __skb_queue_head_init(&pkts);
1504 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1505 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1506 if (unlikely(rc != dlen))
1508 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1509 __skb_queue_purge(&pkts);
1514 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1515 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1516 if (unlikely(rc == -ELINKCONG)) {
1517 tipc_dest_push(clinks, skaddr.node, 0);
1518 tsk->cong_link_cnt++;
1522 if (unlikely(syn && !rc)) {
1523 tipc_set_sk_state(sk, TIPC_CONNECTING);
1524 if (dlen && timeout) {
1525 timeout = msecs_to_jiffies(timeout);
1526 tipc_wait_for_connect(sock, &timeout);
1530 return rc ? rc : dlen;
1534 * tipc_sendstream - send stream-oriented data
1535 * @sock: socket structure
1537 * @dsz: total length of data to be transmitted
1539 * Used for SOCK_STREAM data.
1541 * Return: the number of bytes sent on success (or partial success),
1542 * or errno if no data sent
1544 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1546 struct sock *sk = sock->sk;
1550 ret = __tipc_sendstream(sock, m, dsz);
1556 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1558 struct sock *sk = sock->sk;
1559 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1560 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1561 struct sk_buff_head *txq = &sk->sk_write_queue;
1562 struct tipc_sock *tsk = tipc_sk(sk);
1563 struct tipc_msg *hdr = &tsk->phdr;
1564 struct net *net = sock_net(sk);
1565 struct sk_buff *skb;
1566 u32 dnode = tsk_peer_node(tsk);
1567 int maxnagle = tsk->maxnagle;
1568 int maxpkt = tsk->max_pkt;
1572 if (unlikely(dlen > INT_MAX))
1575 /* Handle implicit connection setup */
1576 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1577 rc = __tipc_sendmsg(sock, m, dlen);
1578 if (dlen && dlen == rc) {
1579 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1580 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1586 rc = tipc_wait_for_cond(sock, &timeout,
1587 (!tsk->cong_link_cnt &&
1588 !tsk_conn_cong(tsk) &&
1589 tipc_sk_connected(sk)));
1592 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1593 blocks = tsk->snd_backlog;
1594 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1596 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1597 if (unlikely(rc < 0))
1601 if (blocks <= 64 && tsk->expect_ack) {
1602 tsk->snd_backlog = blocks;
1605 } else if (blocks > 64) {
1606 tsk->pkt_cnt += skb_queue_len(txq);
1608 skb = skb_peek_tail(txq);
1610 msg_set_ack_required(buf_msg(skb));
1611 tsk->expect_ack = true;
1613 tsk->expect_ack = false;
1619 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1620 if (unlikely(rc != send))
1622 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1624 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1625 TIPC_DUMP_SK_SNDQ, " ");
1626 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1627 if (unlikely(rc == -ELINKCONG)) {
1628 tsk->cong_link_cnt = 1;
1632 tsk->snt_unacked += blocks;
1633 tsk->snd_backlog = 0;
1636 } while (sent < dlen && !rc);
1638 return sent ? sent : rc;
1642 * tipc_send_packet - send a connection-oriented message
1643 * @sock: socket structure
1644 * @m: message to send
1645 * @dsz: length of data to be transmitted
1647 * Used for SOCK_SEQPACKET messages.
1649 * Return: the number of bytes sent on success, or errno otherwise
1651 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1653 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1656 return tipc_sendstream(sock, m, dsz);
1659 /* tipc_sk_finish_conn - complete the setup of a connection
1661 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1664 struct sock *sk = &tsk->sk;
1665 struct net *net = sock_net(sk);
1666 struct tipc_msg *msg = &tsk->phdr;
1668 msg_set_syn(msg, 0);
1669 msg_set_destnode(msg, peer_node);
1670 msg_set_destport(msg, peer_port);
1671 msg_set_type(msg, TIPC_CONN_MSG);
1672 msg_set_lookup_scope(msg, 0);
1673 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1675 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1676 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1677 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1678 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1679 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1681 __skb_queue_purge(&sk->sk_write_queue);
1682 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1685 /* Fall back to message based flow control */
1686 tsk->rcv_win = FLOWCTL_MSG_WIN;
1687 tsk->snd_win = FLOWCTL_MSG_WIN;
1691 * tipc_sk_set_orig_addr - capture sender's address for received message
1692 * @m: descriptor for message info
1693 * @skb: received message
1695 * Note: Address is not captured if not requested by receiver.
1697 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1699 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1700 struct tipc_msg *hdr = buf_msg(skb);
1705 srcaddr->sock.family = AF_TIPC;
1706 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1707 srcaddr->sock.scope = 0;
1708 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1709 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1710 srcaddr->sock.addr.name.domain = 0;
1711 m->msg_namelen = sizeof(struct sockaddr_tipc);
1713 if (!msg_in_group(hdr))
1716 /* Group message users may also want to know sending member's id */
1717 srcaddr->member.family = AF_TIPC;
1718 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1719 srcaddr->member.scope = 0;
1720 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1721 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1722 srcaddr->member.addr.name.domain = 0;
1723 m->msg_namelen = sizeof(*srcaddr);
1727 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1728 * @m: descriptor for message info
1729 * @skb: received message buffer
1730 * @tsk: TIPC port associated with message
1732 * Note: Ancillary data is not captured if not requested by receiver.
1734 * Return: 0 if successful, otherwise errno
1736 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1737 struct tipc_sock *tsk)
1739 struct tipc_msg *hdr;
1744 if (likely(m->msg_controllen == 0))
1748 dlen = msg_data_sz(hdr);
1750 /* Capture errored message object, if any */
1751 if (msg_errcode(hdr)) {
1752 if (skb_linearize(skb))
1755 data[0] = msg_errcode(hdr);
1757 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1760 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1765 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1766 switch (msg_type(hdr)) {
1767 case TIPC_NAMED_MSG:
1769 data[0] = msg_nametype(hdr);
1770 data[1] = msg_namelower(hdr);
1773 case TIPC_MCAST_MSG:
1775 data[0] = msg_nametype(hdr);
1776 data[1] = msg_namelower(hdr);
1777 data[2] = msg_nameupper(hdr);
1780 has_addr = !!tsk->conn_addrtype;
1781 data[0] = msg_nametype(&tsk->phdr);
1782 data[1] = msg_nameinst(&tsk->phdr);
1790 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1793 static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1795 struct sock *sk = &tsk->sk;
1796 struct sk_buff *skb = NULL;
1797 struct tipc_msg *msg;
1798 u32 peer_port = tsk_peer_port(tsk);
1799 u32 dnode = tsk_peer_node(tsk);
1801 if (!tipc_sk_connected(sk))
1803 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1804 dnode, tsk_own_node(tsk), peer_port,
1805 tsk->portid, TIPC_OK);
1809 msg_set_conn_ack(msg, tsk->rcv_unacked);
1810 tsk->rcv_unacked = 0;
1812 /* Adjust to and advertize the correct window limit */
1813 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1814 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1815 msg_set_adv_win(msg, tsk->rcv_win);
1820 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1822 struct sk_buff *skb;
1824 skb = tipc_sk_build_ack(tsk);
1828 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1829 msg_link_selector(buf_msg(skb)));
1832 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1834 struct sock *sk = sock->sk;
1835 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1836 long timeo = *timeop;
1837 int err = sock_error(sk);
1843 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1844 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1848 add_wait_queue(sk_sleep(sk), &wait);
1850 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1851 sched_annotate_sleep();
1853 remove_wait_queue(sk_sleep(sk), &wait);
1856 if (!skb_queue_empty(&sk->sk_receive_queue))
1861 err = sock_intr_errno(timeo);
1862 if (signal_pending(current))
1865 err = sock_error(sk);
1874 * tipc_recvmsg - receive packet-oriented message
1875 * @sock: network socket
1876 * @m: descriptor for message info
1877 * @buflen: length of user buffer area
1878 * @flags: receive flags
1880 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1881 * If the complete message doesn't fit in user area, truncate it.
1883 * Return: size of returned message data, errno otherwise
1885 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1886 size_t buflen, int flags)
1888 struct sock *sk = sock->sk;
1889 bool connected = !tipc_sk_type_connectionless(sk);
1890 struct tipc_sock *tsk = tipc_sk(sk);
1891 int rc, err, hlen, dlen, copy;
1892 struct tipc_skb_cb *skb_cb;
1893 struct sk_buff_head xmitq;
1894 struct tipc_msg *hdr;
1895 struct sk_buff *skb;
1899 /* Catch invalid receive requests */
1900 if (unlikely(!buflen))
1904 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1908 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1910 /* Step rcv queue to first msg with data or error; wait if necessary */
1912 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1915 skb = skb_peek(&sk->sk_receive_queue);
1916 skb_cb = TIPC_SKB_CB(skb);
1918 dlen = msg_data_sz(hdr);
1919 hlen = msg_hdr_sz(hdr);
1920 err = msg_errcode(hdr);
1921 grp_evt = msg_is_grp_evt(hdr);
1922 if (likely(dlen || err))
1924 tsk_advance_rx_queue(sk);
1927 /* Collect msg meta data, including error code and rejected data */
1928 tipc_sk_set_orig_addr(m, skb);
1929 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1934 /* Capture data if non-error msg, otherwise just set return value */
1936 int offset = skb_cb->bytes_read;
1938 copy = min_t(int, dlen - offset, buflen);
1939 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1942 if (unlikely(offset + copy < dlen)) {
1943 if (flags & MSG_EOR) {
1944 if (!(flags & MSG_PEEK))
1945 skb_cb->bytes_read = offset + copy;
1947 m->msg_flags |= MSG_TRUNC;
1948 skb_cb->bytes_read = 0;
1951 if (flags & MSG_EOR)
1952 m->msg_flags |= MSG_EOR;
1953 skb_cb->bytes_read = 0;
1958 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1964 /* Mark message as group event if applicable */
1965 if (unlikely(grp_evt)) {
1966 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1967 m->msg_flags |= MSG_EOR;
1968 m->msg_flags |= MSG_OOB;
1972 /* Caption of data or error code/rejected data was successful */
1973 if (unlikely(flags & MSG_PEEK))
1976 /* Send group flow control advertisement when applicable */
1977 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1978 __skb_queue_head_init(&xmitq);
1979 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1980 msg_orignode(hdr), msg_origport(hdr),
1982 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1985 if (skb_cb->bytes_read)
1988 tsk_advance_rx_queue(sk);
1990 if (likely(!connected))
1993 /* Send connection flow control advertisement when applicable */
1994 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1995 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1996 tipc_sk_send_ack(tsk);
1999 return rc ? rc : copy;
2003 * tipc_recvstream - receive stream-oriented data
2004 * @sock: network socket
2005 * @m: descriptor for message info
2006 * @buflen: total size of user buffer area
2007 * @flags: receive flags
2009 * Used for SOCK_STREAM messages only. If not enough data is available
2010 * will optionally wait for more; never truncates data.
2012 * Return: size of returned message data, errno otherwise
2014 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2015 size_t buflen, int flags)
2017 struct sock *sk = sock->sk;
2018 struct tipc_sock *tsk = tipc_sk(sk);
2019 struct sk_buff *skb;
2020 struct tipc_msg *hdr;
2021 struct tipc_skb_cb *skb_cb;
2022 bool peek = flags & MSG_PEEK;
2023 int offset, required, copy, copied = 0;
2024 int hlen, dlen, err, rc;
2027 /* Catch invalid receive attempts */
2028 if (unlikely(!buflen))
2033 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2037 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2038 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2041 /* Look at first msg in receive queue; wait if necessary */
2042 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2045 skb = skb_peek(&sk->sk_receive_queue);
2046 skb_cb = TIPC_SKB_CB(skb);
2048 dlen = msg_data_sz(hdr);
2049 hlen = msg_hdr_sz(hdr);
2050 err = msg_errcode(hdr);
2052 /* Discard any empty non-errored (SYN-) message */
2053 if (unlikely(!dlen && !err)) {
2054 tsk_advance_rx_queue(sk);
2058 /* Collect msg meta data, incl. error code and rejected data */
2060 tipc_sk_set_orig_addr(m, skb);
2061 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2067 /* Copy data if msg ok, otherwise return error/partial data */
2069 offset = skb_cb->bytes_read;
2070 copy = min_t(int, dlen - offset, buflen - copied);
2071 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2076 if (unlikely(offset < dlen)) {
2078 skb_cb->bytes_read = offset;
2083 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2092 tsk_advance_rx_queue(sk);
2094 /* Send connection flow control advertisement when applicable */
2095 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2096 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2097 tipc_sk_send_ack(tsk);
2099 /* Exit if all requested data or FIN/error received */
2100 if (copied == buflen || err)
2103 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2106 return copied ? copied : rc;
2110 * tipc_write_space - wake up thread if port congestion is released
2113 static void tipc_write_space(struct sock *sk)
2115 struct socket_wq *wq;
2118 wq = rcu_dereference(sk->sk_wq);
2119 if (skwq_has_sleeper(wq))
2120 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2121 EPOLLWRNORM | EPOLLWRBAND);
2126 * tipc_data_ready - wake up threads to indicate messages have been received
2129 static void tipc_data_ready(struct sock *sk)
2131 struct socket_wq *wq;
2134 wq = rcu_dereference(sk->sk_wq);
2135 if (skwq_has_sleeper(wq))
2136 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2137 EPOLLRDNORM | EPOLLRDBAND);
2141 static void tipc_sock_destruct(struct sock *sk)
2143 __skb_queue_purge(&sk->sk_receive_queue);
2146 static void tipc_sk_proto_rcv(struct sock *sk,
2147 struct sk_buff_head *inputq,
2148 struct sk_buff_head *xmitq)
2150 struct sk_buff *skb = __skb_dequeue(inputq);
2151 struct tipc_sock *tsk = tipc_sk(sk);
2152 struct tipc_msg *hdr = buf_msg(skb);
2153 struct tipc_group *grp = tsk->group;
2154 bool wakeup = false;
2156 switch (msg_user(hdr)) {
2158 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2161 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2162 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2164 tsk->cong_link_cnt--;
2166 tipc_sk_push_backlog(tsk, false);
2168 case GROUP_PROTOCOL:
2169 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2172 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2173 hdr, inputq, xmitq);
2180 sk->sk_write_space(sk);
2186 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2188 * @skb: pointer to message buffer.
2189 * @xmitq: for Nagle ACK if any
2190 * Return: true if message should be added to receive queue, false otherwise
2192 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2193 struct sk_buff_head *xmitq)
2195 struct sock *sk = &tsk->sk;
2196 struct net *net = sock_net(sk);
2197 struct tipc_msg *hdr = buf_msg(skb);
2198 bool con_msg = msg_connected(hdr);
2199 u32 pport = tsk_peer_port(tsk);
2200 u32 pnode = tsk_peer_node(tsk);
2201 u32 oport = msg_origport(hdr);
2202 u32 onode = msg_orignode(hdr);
2203 int err = msg_errcode(hdr);
2204 unsigned long delay;
2206 if (unlikely(msg_mcast(hdr)))
2210 switch (sk->sk_state) {
2211 case TIPC_CONNECTING:
2213 if (likely(con_msg)) {
2216 tipc_sk_finish_conn(tsk, oport, onode);
2217 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2218 /* ACK+ message with data is added to receive queue */
2219 if (msg_data_sz(hdr))
2221 /* Empty ACK-, - wake up sleeping connect() and drop */
2222 sk->sk_state_change(sk);
2223 msg_set_dest_droppable(hdr, 1);
2226 /* Ignore connectionless message if not from listening socket */
2227 if (oport != pport || onode != pnode)
2231 if (err != TIPC_ERR_OVERLOAD)
2234 /* Prepare for new setup attempt if we have a SYN clone */
2235 if (skb_queue_empty(&sk->sk_write_queue))
2237 get_random_bytes(&delay, 2);
2238 delay %= (tsk->conn_timeout / 4);
2239 delay = msecs_to_jiffies(delay + 100);
2240 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2243 case TIPC_DISCONNECTING:
2246 /* Accept only SYN message */
2247 if (!msg_is_syn(hdr) &&
2248 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2250 if (!con_msg && !err)
2253 case TIPC_ESTABLISHED:
2254 if (!skb_queue_empty(&sk->sk_write_queue))
2255 tipc_sk_push_backlog(tsk, false);
2256 /* Accept only connection-based messages sent by peer */
2257 if (likely(con_msg && !err && pport == oport &&
2259 if (msg_ack_required(hdr)) {
2260 struct sk_buff *skb;
2262 skb = tipc_sk_build_ack(tsk);
2264 msg_set_nagle_ack(buf_msg(skb));
2265 __skb_queue_tail(xmitq, skb);
2270 if (!tsk_peer_msg(tsk, hdr))
2274 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2275 tipc_node_remove_conn(net, pnode, tsk->portid);
2276 sk->sk_state_change(sk);
2279 pr_err("Unknown sk_state %u\n", sk->sk_state);
2281 /* Abort connection setup attempt */
2282 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2283 sk->sk_err = ECONNREFUSED;
2284 sk->sk_state_change(sk);
2289 * rcvbuf_limit - get proper overload limit of socket receive queue
2293 * For connection oriented messages, irrespective of importance,
2294 * default queue limit is 2 MB.
2296 * For connectionless messages, queue limits are based on message
2297 * importance as follows:
2299 * TIPC_LOW_IMPORTANCE (2 MB)
2300 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2301 * TIPC_HIGH_IMPORTANCE (8 MB)
2302 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2304 * Return: overload limit according to corresponding message importance
2306 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2308 struct tipc_sock *tsk = tipc_sk(sk);
2309 struct tipc_msg *hdr = buf_msg(skb);
2311 if (unlikely(msg_in_group(hdr)))
2312 return READ_ONCE(sk->sk_rcvbuf);
2314 if (unlikely(!msg_connected(hdr)))
2315 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2317 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2318 return READ_ONCE(sk->sk_rcvbuf);
2320 return FLOWCTL_MSG_LIM;
2324 * tipc_sk_filter_rcv - validate incoming message
2326 * @skb: pointer to message.
2327 * @xmitq: output message area (FIXME)
2329 * Enqueues message on receive queue if acceptable; optionally handles
2330 * disconnect indication for a connected socket.
2332 * Called with socket lock already taken
2334 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2335 struct sk_buff_head *xmitq)
2337 bool sk_conn = !tipc_sk_type_connectionless(sk);
2338 struct tipc_sock *tsk = tipc_sk(sk);
2339 struct tipc_group *grp = tsk->group;
2340 struct tipc_msg *hdr = buf_msg(skb);
2341 struct net *net = sock_net(sk);
2342 struct sk_buff_head inputq;
2343 int mtyp = msg_type(hdr);
2344 int limit, err = TIPC_OK;
2346 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2347 TIPC_SKB_CB(skb)->bytes_read = 0;
2348 __skb_queue_head_init(&inputq);
2349 __skb_queue_tail(&inputq, skb);
2351 if (unlikely(!msg_isdata(hdr)))
2352 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2355 tipc_group_filter_msg(grp, &inputq, xmitq);
2357 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2358 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2360 /* Validate and add to receive buffer if there is space */
2361 while ((skb = __skb_dequeue(&inputq))) {
2363 limit = rcvbuf_limit(sk, skb);
2364 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2365 (!sk_conn && msg_connected(hdr)) ||
2366 (!grp && msg_in_group(hdr)))
2367 err = TIPC_ERR_NO_PORT;
2368 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2369 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2371 atomic_inc(&sk->sk_drops);
2372 err = TIPC_ERR_OVERLOAD;
2375 if (unlikely(err)) {
2376 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2377 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2379 __skb_queue_tail(xmitq, skb);
2384 __skb_queue_tail(&sk->sk_receive_queue, skb);
2385 skb_set_owner_r(skb, sk);
2386 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2387 "rcvq >90% allocated!");
2388 sk->sk_data_ready(sk);
2393 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2397 * Caller must hold socket lock
2399 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2401 unsigned int before = sk_rmem_alloc_get(sk);
2402 struct sk_buff_head xmitq;
2405 __skb_queue_head_init(&xmitq);
2407 tipc_sk_filter_rcv(sk, skb, &xmitq);
2408 added = sk_rmem_alloc_get(sk) - before;
2409 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2411 /* Send pending response/rejected messages, if any */
2412 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2417 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2418 * inputq and try adding them to socket or backlog queue
2419 * @inputq: list of incoming buffers with potentially different destinations
2420 * @sk: socket where the buffers should be enqueued
2421 * @dport: port number for the socket
2422 * @xmitq: output queue
2424 * Caller must hold socket lock
2426 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2427 u32 dport, struct sk_buff_head *xmitq)
2429 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2430 struct sk_buff *skb;
2435 while (skb_queue_len(inputq)) {
2436 if (unlikely(time_after_eq(jiffies, time_limit)))
2439 skb = tipc_skb_dequeue(inputq, dport);
2443 /* Add message directly to receive queue if possible */
2444 if (!sock_owned_by_user(sk)) {
2445 tipc_sk_filter_rcv(sk, skb, xmitq);
2449 /* Try backlog, compensating for double-counted bytes */
2450 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2451 if (!sk->sk_backlog.len)
2452 atomic_set(dcnt, 0);
2453 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2454 if (likely(!sk_add_backlog(sk, skb, lim))) {
2455 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2456 "bklg & rcvq >90% allocated!");
2460 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2461 /* Overload => reject message back to sender */
2462 onode = tipc_own_addr(sock_net(sk));
2463 atomic_inc(&sk->sk_drops);
2464 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2465 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2467 __skb_queue_tail(xmitq, skb);
2474 * tipc_sk_rcv - handle a chain of incoming buffers
2475 * @net: the associated network namespace
2476 * @inputq: buffer list containing the buffers
2477 * Consumes all buffers in list until inputq is empty
2478 * Note: may be called in multiple threads referring to the same queue
2480 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2482 struct sk_buff_head xmitq;
2483 u32 dnode, dport = 0;
2485 struct tipc_sock *tsk;
2487 struct sk_buff *skb;
2489 __skb_queue_head_init(&xmitq);
2490 while (skb_queue_len(inputq)) {
2491 dport = tipc_skb_peek_port(inputq, dport);
2492 tsk = tipc_sk_lookup(net, dport);
2496 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2497 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2498 spin_unlock_bh(&sk->sk_lock.slock);
2500 /* Send pending response/rejected messages, if any */
2501 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2505 /* No destination socket => dequeue skb if still there */
2506 skb = tipc_skb_dequeue(inputq, dport);
2510 /* Try secondary lookup if unresolved named message */
2511 err = TIPC_ERR_NO_PORT;
2512 if (tipc_msg_lookup_dest(net, skb, &err))
2515 /* Prepare for message rejection */
2516 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2519 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2521 dnode = msg_destnode(buf_msg(skb));
2522 tipc_node_xmit_skb(net, skb, dnode, dport);
2526 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2528 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2529 struct sock *sk = sock->sk;
2533 int err = sock_error(sk);
2538 if (signal_pending(current))
2539 return sock_intr_errno(*timeo_p);
2540 if (sk->sk_state == TIPC_DISCONNECTING)
2543 add_wait_queue(sk_sleep(sk), &wait);
2544 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2546 remove_wait_queue(sk_sleep(sk), &wait);
2551 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2553 if (addr->family != AF_TIPC)
2555 if (addr->addrtype == TIPC_SERVICE_RANGE)
2556 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2557 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2558 addr->addrtype == TIPC_SOCKET_ADDR);
2562 * tipc_connect - establish a connection to another TIPC port
2563 * @sock: socket structure
2564 * @dest: socket address for destination port
2565 * @destlen: size of socket address data structure
2566 * @flags: file-related flags associated with socket
2568 * Return: 0 on success, errno otherwise
2570 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2571 int destlen, int flags)
2573 struct sock *sk = sock->sk;
2574 struct tipc_sock *tsk = tipc_sk(sk);
2575 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2576 struct msghdr m = {NULL,};
2577 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2581 if (destlen != sizeof(struct sockaddr_tipc))
2591 if (dst->family == AF_UNSPEC) {
2592 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2593 if (!tipc_sk_type_connectionless(sk))
2597 if (!tipc_sockaddr_is_sane(dst)) {
2601 /* DGRAM/RDM connect(), just save the destaddr */
2602 if (tipc_sk_type_connectionless(sk)) {
2603 memcpy(&tsk->peer, dest, destlen);
2605 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2610 previous = sk->sk_state;
2612 switch (sk->sk_state) {
2614 /* Send a 'SYN-' to destination */
2616 m.msg_namelen = destlen;
2618 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2619 * indicate send_msg() is never blocked.
2622 m.msg_flags = MSG_DONTWAIT;
2624 res = __tipc_sendmsg(sock, &m, 0);
2625 if ((res < 0) && (res != -EWOULDBLOCK))
2628 /* Just entered TIPC_CONNECTING state; the only
2629 * difference is that return value in non-blocking
2630 * case is EINPROGRESS, rather than EALREADY.
2634 case TIPC_CONNECTING:
2636 if (previous == TIPC_CONNECTING)
2640 timeout = msecs_to_jiffies(timeout);
2641 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2642 res = tipc_wait_for_connect(sock, &timeout);
2644 case TIPC_ESTABLISHED:
2657 * tipc_listen - allow socket to listen for incoming connections
2658 * @sock: socket structure
2661 * Return: 0 on success, errno otherwise
2663 static int tipc_listen(struct socket *sock, int len)
2665 struct sock *sk = sock->sk;
2669 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2675 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2677 struct sock *sk = sock->sk;
2678 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2681 /* True wake-one mechanism for incoming connections: only
2682 * one process gets woken up, not the 'whole herd'.
2683 * Since we do not 'race & poll' for established sockets
2684 * anymore, the common case will execute the loop only once.
2687 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2688 add_wait_queue(sk_sleep(sk), &wait);
2690 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2692 remove_wait_queue(sk_sleep(sk), &wait);
2695 if (!skb_queue_empty(&sk->sk_receive_queue))
2700 err = sock_intr_errno(timeo);
2701 if (signal_pending(current))
2708 * tipc_accept - wait for connection request
2709 * @sock: listening socket
2710 * @new_sock: new socket that is to be connected
2711 * @flags: file-related flags associated with socket
2712 * @kern: caused by kernel or by userspace?
2714 * Return: 0 on success, errno otherwise
2716 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2719 struct sock *new_sk, *sk = sock->sk;
2720 struct tipc_sock *new_tsock;
2721 struct msghdr m = {NULL,};
2722 struct tipc_msg *msg;
2723 struct sk_buff *buf;
2729 if (sk->sk_state != TIPC_LISTEN) {
2733 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2734 res = tipc_wait_for_accept(sock, timeo);
2738 buf = skb_peek(&sk->sk_receive_queue);
2740 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2743 security_sk_clone(sock->sk, new_sock->sk);
2745 new_sk = new_sock->sk;
2746 new_tsock = tipc_sk(new_sk);
2749 /* we lock on new_sk; but lockdep sees the lock on sk */
2750 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2753 * Reject any stray messages received by new socket
2754 * before the socket lock was taken (very, very unlikely)
2756 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2758 /* Connect new socket to it's peer */
2759 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2761 tsk_set_importance(new_sk, msg_importance(msg));
2762 if (msg_named(msg)) {
2763 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2764 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2765 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2769 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2770 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2772 if (!msg_data_sz(msg)) {
2773 tsk_advance_rx_queue(sk);
2775 __skb_dequeue(&sk->sk_receive_queue);
2776 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2777 skb_set_owner_r(buf, new_sk);
2779 __tipc_sendstream(new_sock, &m, 0);
2780 release_sock(new_sk);
2787 * tipc_shutdown - shutdown socket connection
2788 * @sock: socket structure
2789 * @how: direction to close (must be SHUT_RDWR)
2791 * Terminates connection (if necessary), then purges socket's receive queue.
2793 * Return: 0 on success, errno otherwise
2795 static int tipc_shutdown(struct socket *sock, int how)
2797 struct sock *sk = sock->sk;
2800 if (how != SHUT_RDWR)
2805 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2806 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2807 sk->sk_shutdown = SHUTDOWN_MASK;
2809 if (sk->sk_state == TIPC_DISCONNECTING) {
2810 /* Discard any unreceived messages */
2811 __skb_queue_purge(&sk->sk_receive_queue);
2817 /* Wake up anyone sleeping in poll. */
2818 sk->sk_state_change(sk);
2824 static void tipc_sk_check_probing_state(struct sock *sk,
2825 struct sk_buff_head *list)
2827 struct tipc_sock *tsk = tipc_sk(sk);
2828 u32 pnode = tsk_peer_node(tsk);
2829 u32 pport = tsk_peer_port(tsk);
2830 u32 self = tsk_own_node(tsk);
2831 u32 oport = tsk->portid;
2832 struct sk_buff *skb;
2834 if (tsk->probe_unacked) {
2835 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2836 sk->sk_err = ECONNABORTED;
2837 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2838 sk->sk_state_change(sk);
2841 /* Prepare new probe */
2842 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2843 pnode, self, pport, oport, TIPC_OK);
2845 __skb_queue_tail(list, skb);
2846 tsk->probe_unacked = true;
2847 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2850 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2852 struct tipc_sock *tsk = tipc_sk(sk);
2854 /* Try again later if dest link is congested */
2855 if (tsk->cong_link_cnt) {
2856 sk_reset_timer(sk, &sk->sk_timer,
2857 jiffies + msecs_to_jiffies(100));
2860 /* Prepare SYN for retransmit */
2861 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2864 static void tipc_sk_timeout(struct timer_list *t)
2866 struct sock *sk = from_timer(sk, t, sk_timer);
2867 struct tipc_sock *tsk = tipc_sk(sk);
2868 u32 pnode = tsk_peer_node(tsk);
2869 struct sk_buff_head list;
2872 __skb_queue_head_init(&list);
2875 /* Try again later if socket is busy */
2876 if (sock_owned_by_user(sk)) {
2877 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2883 if (sk->sk_state == TIPC_ESTABLISHED)
2884 tipc_sk_check_probing_state(sk, &list);
2885 else if (sk->sk_state == TIPC_CONNECTING)
2886 tipc_sk_retry_connect(sk, &list);
2890 if (!skb_queue_empty(&list))
2891 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2893 /* SYN messages may cause link congestion */
2894 if (rc == -ELINKCONG) {
2895 tipc_dest_push(&tsk->cong_links, pnode, 0);
2896 tsk->cong_link_cnt = 1;
2901 static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2903 struct sock *sk = &tsk->sk;
2904 struct net *net = sock_net(sk);
2905 struct tipc_socket_addr skaddr;
2906 struct publication *p;
2909 if (tipc_sk_connected(sk))
2911 key = tsk->portid + tsk->pub_count + 1;
2912 if (key == tsk->portid)
2914 skaddr.ref = tsk->portid;
2915 skaddr.node = tipc_own_addr(net);
2916 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2920 list_add(&p->binding_sock, &tsk->publications);
2922 tsk->published = true;
2926 static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2928 struct net *net = sock_net(&tsk->sk);
2929 struct publication *safe, *p;
2930 struct tipc_uaddr _ua;
2933 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2935 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2936 p->sr.type, p->sr.lower, p->sr.upper);
2937 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2940 /* Unbind specific publication */
2941 if (p->scope != ua->scope)
2943 if (p->sr.type != ua->sr.type)
2945 if (p->sr.lower != ua->sr.lower)
2947 if (p->sr.upper != ua->sr.upper)
2949 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2953 if (list_empty(&tsk->publications)) {
2960 /* tipc_sk_reinit: set non-zero address in all existing sockets
2961 * when we go from standalone to network mode.
2963 void tipc_sk_reinit(struct net *net)
2965 struct tipc_net *tn = net_generic(net, tipc_net_id);
2966 struct rhashtable_iter iter;
2967 struct tipc_sock *tsk;
2968 struct tipc_msg *msg;
2970 rhashtable_walk_enter(&tn->sk_rht, &iter);
2973 rhashtable_walk_start(&iter);
2975 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2976 sock_hold(&tsk->sk);
2977 rhashtable_walk_stop(&iter);
2978 lock_sock(&tsk->sk);
2980 msg_set_prevnode(msg, tipc_own_addr(net));
2981 msg_set_orignode(msg, tipc_own_addr(net));
2982 release_sock(&tsk->sk);
2983 rhashtable_walk_start(&iter);
2987 rhashtable_walk_stop(&iter);
2988 } while (tsk == ERR_PTR(-EAGAIN));
2990 rhashtable_walk_exit(&iter);
2993 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2995 struct tipc_net *tn = net_generic(net, tipc_net_id);
2996 struct tipc_sock *tsk;
2999 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
3001 sock_hold(&tsk->sk);
3007 static int tipc_sk_insert(struct tipc_sock *tsk)
3009 struct sock *sk = &tsk->sk;
3010 struct net *net = sock_net(sk);
3011 struct tipc_net *tn = net_generic(net, tipc_net_id);
3012 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3013 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3015 while (remaining--) {
3017 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3018 portid = TIPC_MIN_PORT;
3019 tsk->portid = portid;
3020 sock_hold(&tsk->sk);
3021 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3030 static void tipc_sk_remove(struct tipc_sock *tsk)
3032 struct sock *sk = &tsk->sk;
3033 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3035 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3036 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3041 static const struct rhashtable_params tsk_rht_params = {
3043 .head_offset = offsetof(struct tipc_sock, node),
3044 .key_offset = offsetof(struct tipc_sock, portid),
3045 .key_len = sizeof(u32), /* portid */
3046 .max_size = 1048576,
3048 .automatic_shrinking = true,
3051 int tipc_sk_rht_init(struct net *net)
3053 struct tipc_net *tn = net_generic(net, tipc_net_id);
3055 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3058 void tipc_sk_rht_destroy(struct net *net)
3060 struct tipc_net *tn = net_generic(net, tipc_net_id);
3062 /* Wait for socket readers to complete */
3065 rhashtable_destroy(&tn->sk_rht);
3068 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3070 struct net *net = sock_net(&tsk->sk);
3071 struct tipc_group *grp = tsk->group;
3072 struct tipc_msg *hdr = &tsk->phdr;
3073 struct tipc_uaddr ua;
3076 if (mreq->type < TIPC_RESERVED_TYPES)
3078 if (mreq->scope > TIPC_NODE_SCOPE)
3080 if (mreq->scope != TIPC_NODE_SCOPE)
3081 mreq->scope = TIPC_CLUSTER_SCOPE;
3084 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3088 msg_set_lookup_scope(hdr, mreq->scope);
3089 msg_set_nametype(hdr, mreq->type);
3090 msg_set_dest_droppable(hdr, true);
3091 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3092 mreq->type, mreq->instance, mreq->instance);
3093 tipc_nametbl_build_group(net, grp, &ua);
3094 rc = tipc_sk_publish(tsk, &ua);
3096 tipc_group_delete(net, grp);
3100 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3101 tsk->mc_method.rcast = true;
3102 tsk->mc_method.mandatory = true;
3103 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3107 static int tipc_sk_leave(struct tipc_sock *tsk)
3109 struct net *net = sock_net(&tsk->sk);
3110 struct tipc_group *grp = tsk->group;
3111 struct tipc_uaddr ua;
3116 ua.addrtype = TIPC_SERVICE_RANGE;
3117 tipc_group_self(grp, &ua.sr, &scope);
3119 tipc_group_delete(net, grp);
3121 tipc_sk_withdraw(tsk, &ua);
3126 * tipc_setsockopt - set socket option
3127 * @sock: socket structure
3128 * @lvl: option level
3129 * @opt: option identifier
3130 * @ov: pointer to new option value
3131 * @ol: length of option value
3133 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3134 * (to ease compatibility).
3136 * Return: 0 on success, errno otherwise
3138 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3139 sockptr_t ov, unsigned int ol)
3141 struct sock *sk = sock->sk;
3142 struct tipc_sock *tsk = tipc_sk(sk);
3143 struct tipc_group_req mreq;
3147 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3149 if (lvl != SOL_TIPC)
3150 return -ENOPROTOOPT;
3153 case TIPC_IMPORTANCE:
3154 case TIPC_SRC_DROPPABLE:
3155 case TIPC_DEST_DROPPABLE:
3156 case TIPC_CONN_TIMEOUT:
3158 if (ol < sizeof(value))
3160 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3163 case TIPC_GROUP_JOIN:
3164 if (ol < sizeof(mreq))
3166 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3170 if (!sockptr_is_null(ov) || ol)
3177 case TIPC_IMPORTANCE:
3178 res = tsk_set_importance(sk, value);
3180 case TIPC_SRC_DROPPABLE:
3181 if (sock->type != SOCK_STREAM)
3182 tsk_set_unreliable(tsk, value);
3186 case TIPC_DEST_DROPPABLE:
3187 tsk_set_unreturnable(tsk, value);
3189 case TIPC_CONN_TIMEOUT:
3190 tipc_sk(sk)->conn_timeout = value;
3192 case TIPC_MCAST_BROADCAST:
3193 tsk->mc_method.rcast = false;
3194 tsk->mc_method.mandatory = true;
3196 case TIPC_MCAST_REPLICAST:
3197 tsk->mc_method.rcast = true;
3198 tsk->mc_method.mandatory = true;
3200 case TIPC_GROUP_JOIN:
3201 res = tipc_sk_join(tsk, &mreq);
3203 case TIPC_GROUP_LEAVE:
3204 res = tipc_sk_leave(tsk);
3207 tsk->nodelay = !!value;
3220 * tipc_getsockopt - get socket option
3221 * @sock: socket structure
3222 * @lvl: option level
3223 * @opt: option identifier
3224 * @ov: receptacle for option value
3225 * @ol: receptacle for length of option value
3227 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3228 * (to ease compatibility).
3230 * Return: 0 on success, errno otherwise
3232 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3233 char __user *ov, int __user *ol)
3235 struct sock *sk = sock->sk;
3236 struct tipc_sock *tsk = tipc_sk(sk);
3237 struct tipc_service_range seq;
3242 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3243 return put_user(0, ol);
3244 if (lvl != SOL_TIPC)
3245 return -ENOPROTOOPT;
3246 res = get_user(len, ol);
3253 case TIPC_IMPORTANCE:
3254 value = tsk_importance(tsk);
3256 case TIPC_SRC_DROPPABLE:
3257 value = tsk_unreliable(tsk);
3259 case TIPC_DEST_DROPPABLE:
3260 value = tsk_unreturnable(tsk);
3262 case TIPC_CONN_TIMEOUT:
3263 value = tsk->conn_timeout;
3264 /* no need to set "res", since already 0 at this point */
3266 case TIPC_NODE_RECVQ_DEPTH:
3267 value = 0; /* was tipc_queue_size, now obsolete */
3269 case TIPC_SOCK_RECVQ_DEPTH:
3270 value = skb_queue_len(&sk->sk_receive_queue);
3272 case TIPC_SOCK_RECVQ_USED:
3273 value = sk_rmem_alloc_get(sk);
3275 case TIPC_GROUP_JOIN:
3278 tipc_group_self(tsk->group, &seq, &scope);
3288 return res; /* "get" failed */
3290 if (len < sizeof(value))
3293 if (copy_to_user(ov, &value, sizeof(value)))
3296 return put_user(sizeof(value), ol);
3299 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3301 struct net *net = sock_net(sock->sk);
3302 struct tipc_sioc_nodeid_req nr = {0};
3303 struct tipc_sioc_ln_req lnr;
3304 void __user *argp = (void __user *)arg;
3307 case SIOCGETLINKNAME:
3308 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3310 if (!tipc_node_get_linkname(net,
3311 lnr.bearer_id & 0xffff, lnr.peer,
3312 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3313 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3317 return -EADDRNOTAVAIL;
3319 if (copy_from_user(&nr, argp, sizeof(nr)))
3321 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3322 return -EADDRNOTAVAIL;
3323 if (copy_to_user(argp, &nr, sizeof(nr)))
3327 return -ENOIOCTLCMD;
3331 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3333 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3334 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3335 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3337 tsk1->peer.family = AF_TIPC;
3338 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3339 tsk1->peer.scope = TIPC_NODE_SCOPE;
3340 tsk1->peer.addr.id.ref = tsk2->portid;
3341 tsk1->peer.addr.id.node = onode;
3342 tsk2->peer.family = AF_TIPC;
3343 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3344 tsk2->peer.scope = TIPC_NODE_SCOPE;
3345 tsk2->peer.addr.id.ref = tsk1->portid;
3346 tsk2->peer.addr.id.node = onode;
3348 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3349 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3353 /* Protocol switches for the various types of TIPC sockets */
3355 static const struct proto_ops msg_ops = {
3356 .owner = THIS_MODULE,
3358 .release = tipc_release,
3360 .connect = tipc_connect,
3361 .socketpair = tipc_socketpair,
3362 .accept = sock_no_accept,
3363 .getname = tipc_getname,
3365 .ioctl = tipc_ioctl,
3366 .listen = sock_no_listen,
3367 .shutdown = tipc_shutdown,
3368 .setsockopt = tipc_setsockopt,
3369 .getsockopt = tipc_getsockopt,
3370 .sendmsg = tipc_sendmsg,
3371 .recvmsg = tipc_recvmsg,
3372 .mmap = sock_no_mmap,
3373 .sendpage = sock_no_sendpage
3376 static const struct proto_ops packet_ops = {
3377 .owner = THIS_MODULE,
3379 .release = tipc_release,
3381 .connect = tipc_connect,
3382 .socketpair = tipc_socketpair,
3383 .accept = tipc_accept,
3384 .getname = tipc_getname,
3386 .ioctl = tipc_ioctl,
3387 .listen = tipc_listen,
3388 .shutdown = tipc_shutdown,
3389 .setsockopt = tipc_setsockopt,
3390 .getsockopt = tipc_getsockopt,
3391 .sendmsg = tipc_send_packet,
3392 .recvmsg = tipc_recvmsg,
3393 .mmap = sock_no_mmap,
3394 .sendpage = sock_no_sendpage
3397 static const struct proto_ops stream_ops = {
3398 .owner = THIS_MODULE,
3400 .release = tipc_release,
3402 .connect = tipc_connect,
3403 .socketpair = tipc_socketpair,
3404 .accept = tipc_accept,
3405 .getname = tipc_getname,
3407 .ioctl = tipc_ioctl,
3408 .listen = tipc_listen,
3409 .shutdown = tipc_shutdown,
3410 .setsockopt = tipc_setsockopt,
3411 .getsockopt = tipc_getsockopt,
3412 .sendmsg = tipc_sendstream,
3413 .recvmsg = tipc_recvstream,
3414 .mmap = sock_no_mmap,
3415 .sendpage = sock_no_sendpage
3418 static const struct net_proto_family tipc_family_ops = {
3419 .owner = THIS_MODULE,
3421 .create = tipc_sk_create
3424 static struct proto tipc_proto = {
3426 .owner = THIS_MODULE,
3427 .obj_size = sizeof(struct tipc_sock),
3428 .sysctl_rmem = sysctl_tipc_rmem
3432 * tipc_socket_init - initialize TIPC socket interface
3434 * Return: 0 on success, errno otherwise
3436 int tipc_socket_init(void)
3440 res = proto_register(&tipc_proto, 1);
3442 pr_err("Failed to register TIPC protocol type\n");
3446 res = sock_register(&tipc_family_ops);
3448 pr_err("Failed to register TIPC socket type\n");
3449 proto_unregister(&tipc_proto);
3457 * tipc_socket_stop - stop TIPC socket interface
3459 void tipc_socket_stop(void)
3461 sock_unregister(tipc_family_ops.family);
3462 proto_unregister(&tipc_proto);
3465 /* Caller should hold socket lock for the passed tipc socket. */
3466 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3468 u32 peer_node, peer_port;
3469 u32 conn_type, conn_instance;
3470 struct nlattr *nest;
3472 peer_node = tsk_peer_node(tsk);
3473 peer_port = tsk_peer_port(tsk);
3474 conn_type = msg_nametype(&tsk->phdr);
3475 conn_instance = msg_nameinst(&tsk->phdr);
3476 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3480 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3482 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3485 if (tsk->conn_addrtype != 0) {
3486 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3488 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3490 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3493 nla_nest_end(skb, nest);
3498 nla_nest_cancel(skb, nest);
3503 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3506 struct net *net = sock_net(skb->sk);
3507 struct sock *sk = &tsk->sk;
3509 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3510 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3513 if (tipc_sk_connected(sk)) {
3514 if (__tipc_nl_add_sk_con(skb, tsk))
3516 } else if (!list_empty(&tsk->publications)) {
3517 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3523 /* Caller should hold socket lock for the passed tipc socket. */
3524 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3525 struct tipc_sock *tsk)
3527 struct nlattr *attrs;
3530 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3531 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3535 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3537 goto genlmsg_cancel;
3539 if (__tipc_nl_add_sk_info(skb, tsk))
3540 goto attr_msg_cancel;
3542 nla_nest_end(skb, attrs);
3543 genlmsg_end(skb, hdr);
3548 nla_nest_cancel(skb, attrs);
3550 genlmsg_cancel(skb, hdr);
3555 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3556 int (*skb_handler)(struct sk_buff *skb,
3557 struct netlink_callback *cb,
3558 struct tipc_sock *tsk))
3560 struct rhashtable_iter *iter = (void *)cb->args[4];
3561 struct tipc_sock *tsk;
3564 rhashtable_walk_start(iter);
3565 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3568 if (err == -EAGAIN) {
3575 sock_hold(&tsk->sk);
3576 rhashtable_walk_stop(iter);
3577 lock_sock(&tsk->sk);
3578 err = skb_handler(skb, cb, tsk);
3580 release_sock(&tsk->sk);
3584 release_sock(&tsk->sk);
3585 rhashtable_walk_start(iter);
3588 rhashtable_walk_stop(iter);
3592 EXPORT_SYMBOL(tipc_nl_sk_walk);
3594 int tipc_dump_start(struct netlink_callback *cb)
3596 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3598 EXPORT_SYMBOL(tipc_dump_start);
3600 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3602 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3603 struct rhashtable_iter *iter = (void *)cb->args[4];
3604 struct tipc_net *tn = tipc_net(net);
3607 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3611 cb->args[4] = (long)iter;
3614 rhashtable_walk_enter(&tn->sk_rht, iter);
3618 int tipc_dump_done(struct netlink_callback *cb)
3620 struct rhashtable_iter *hti = (void *)cb->args[4];
3622 rhashtable_walk_exit(hti);
3626 EXPORT_SYMBOL(tipc_dump_done);
3628 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3629 struct tipc_sock *tsk, u32 sk_filter_state,
3630 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3632 struct sock *sk = &tsk->sk;
3633 struct nlattr *attrs;
3634 struct nlattr *stat;
3636 /*filter response w.r.t sk_state*/
3637 if (!(sk_filter_state & (1 << sk->sk_state)))
3640 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3644 if (__tipc_nl_add_sk_info(skb, tsk))
3645 goto attr_msg_cancel;
3647 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3648 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3649 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3650 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3651 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3653 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3654 tipc_diag_gen_cookie(sk),
3656 goto attr_msg_cancel;
3658 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3660 goto attr_msg_cancel;
3662 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3663 skb_queue_len(&sk->sk_receive_queue)) ||
3664 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3665 skb_queue_len(&sk->sk_write_queue)) ||
3666 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3667 atomic_read(&sk->sk_drops)))
3668 goto stat_msg_cancel;
3670 if (tsk->cong_link_cnt &&
3671 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3672 goto stat_msg_cancel;
3674 if (tsk_conn_cong(tsk) &&
3675 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3676 goto stat_msg_cancel;
3678 nla_nest_end(skb, stat);
3681 if (tipc_group_fill_sock_diag(tsk->group, skb))
3682 goto stat_msg_cancel;
3684 nla_nest_end(skb, attrs);
3689 nla_nest_cancel(skb, stat);
3691 nla_nest_cancel(skb, attrs);
3695 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3697 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3699 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3702 /* Caller should hold socket lock for the passed tipc socket. */
3703 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3704 struct netlink_callback *cb,
3705 struct publication *publ)
3708 struct nlattr *attrs;
3710 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3711 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3715 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3717 goto genlmsg_cancel;
3719 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3720 goto attr_msg_cancel;
3721 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3722 goto attr_msg_cancel;
3723 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3724 goto attr_msg_cancel;
3725 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3726 goto attr_msg_cancel;
3728 nla_nest_end(skb, attrs);
3729 genlmsg_end(skb, hdr);
3734 nla_nest_cancel(skb, attrs);
3736 genlmsg_cancel(skb, hdr);
3741 /* Caller should hold socket lock for the passed tipc socket. */
3742 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3743 struct netlink_callback *cb,
3744 struct tipc_sock *tsk, u32 *last_publ)
3747 struct publication *p;
3750 list_for_each_entry(p, &tsk->publications, binding_sock) {
3751 if (p->key == *last_publ)
3754 if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3755 /* We never set seq or call nl_dump_check_consistent()
3756 * this means that setting prev_seq here will cause the
3757 * consistence check to fail in the netlink callback
3758 * handler. Resulting in the last NLMSG_DONE message
3759 * having the NLM_F_DUMP_INTR flag set.
3766 p = list_first_entry(&tsk->publications, struct publication,
3770 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3771 err = __tipc_nl_add_sk_publ(skb, cb, p);
3773 *last_publ = p->key;
3782 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3785 u32 tsk_portid = cb->args[0];
3786 u32 last_publ = cb->args[1];
3787 u32 done = cb->args[2];
3788 struct net *net = sock_net(skb->sk);
3789 struct tipc_sock *tsk;
3792 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3793 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3795 if (!attrs[TIPC_NLA_SOCK])
3798 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3799 attrs[TIPC_NLA_SOCK],
3800 tipc_nl_sock_policy, NULL);
3804 if (!sock[TIPC_NLA_SOCK_REF])
3807 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3813 tsk = tipc_sk_lookup(net, tsk_portid);
3817 lock_sock(&tsk->sk);
3818 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3821 release_sock(&tsk->sk);
3824 cb->args[0] = tsk_portid;
3825 cb->args[1] = last_publ;
3832 * tipc_sk_filtering - check if a socket should be traced
3833 * @sk: the socket to be examined
3835 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3836 * (portid, sock type, name type, name lower, name upper)
3838 * Return: true if the socket meets the socket tuple data
3839 * (value 0 = 'any') or when there is no tuple set (all = 0),
3842 bool tipc_sk_filtering(struct sock *sk)
3844 struct tipc_sock *tsk;
3845 struct publication *p;
3846 u32 _port, _sktype, _type, _lower, _upper;
3847 u32 type = 0, lower = 0, upper = 0;
3854 _port = sysctl_tipc_sk_filter[0];
3855 _sktype = sysctl_tipc_sk_filter[1];
3856 _type = sysctl_tipc_sk_filter[2];
3857 _lower = sysctl_tipc_sk_filter[3];
3858 _upper = sysctl_tipc_sk_filter[4];
3860 if (!_port && !_sktype && !_type && !_lower && !_upper)
3864 return (_port == tsk->portid);
3866 if (_sktype && _sktype != sk->sk_type)
3869 if (tsk->published) {
3870 p = list_first_entry_or_null(&tsk->publications,
3871 struct publication, binding_sock);
3874 lower = p->sr.lower;
3875 upper = p->sr.upper;
3879 if (!tipc_sk_type_connectionless(sk)) {
3880 type = msg_nametype(&tsk->phdr);
3881 lower = msg_nameinst(&tsk->phdr);
3885 if ((_type && _type != type) || (_lower && _lower != lower) ||
3886 (_upper && _upper != upper))
3892 u32 tipc_sock_get_portid(struct sock *sk)
3894 return (sk) ? (tipc_sk(sk))->portid : 0;
3898 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3899 * both the rcv and backlog queues are considered
3900 * @sk: tipc sk to be checked
3901 * @skb: tipc msg to be checked
3903 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3906 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3908 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3909 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3910 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3912 return (qsize > lim * 90 / 100);
3916 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3917 * only the rcv queue is considered
3918 * @sk: tipc sk to be checked
3919 * @skb: tipc msg to be checked
3921 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3924 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3926 unsigned int lim = rcvbuf_limit(sk, skb);
3927 unsigned int qsize = sk_rmem_alloc_get(sk);
3929 return (qsize > lim * 90 / 100);
3933 * tipc_sk_dump - dump TIPC socket
3934 * @sk: tipc sk to be dumped
3935 * @dqueues: bitmask to decide if any socket queue to be dumped?
3936 * - TIPC_DUMP_NONE: don't dump socket queues
3937 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3938 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3939 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3940 * - TIPC_DUMP_ALL: dump all the socket queues above
3941 * @buf: returned buffer of dump data in format
3943 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3946 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3947 u32 conn_type, conn_instance;
3948 struct tipc_sock *tsk;
3949 struct publication *p;
3953 i += scnprintf(buf, sz, "sk data: (null)\n");
3958 tsk_connected = !tipc_sk_type_connectionless(sk);
3960 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3961 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3962 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3963 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3964 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3965 if (tsk_connected) {
3966 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3967 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3968 conn_type = msg_nametype(&tsk->phdr);
3969 conn_instance = msg_nameinst(&tsk->phdr);
3970 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3971 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3973 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3974 if (tsk->published) {
3975 p = list_first_entry_or_null(&tsk->publications,
3976 struct publication, binding_sock);
3977 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3978 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3979 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3981 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3982 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3983 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3984 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3985 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3986 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3987 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3988 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3989 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3990 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3991 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3992 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3993 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3994 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3996 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3997 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3998 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
4001 if (dqueues & TIPC_DUMP_SK_RCVQ) {
4002 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
4003 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4006 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4007 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4008 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4009 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4010 i += scnprintf(buf + i, sz - i, " tail ");
4011 i += tipc_skb_dump(sk->sk_backlog.tail, false,