2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2017, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include <linux/rhashtable.h>
38 #include <linux/sched/signal.h>
41 #include "name_table.h"
44 #include "name_distr.h"
51 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
52 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
53 #define TIPC_FWD_MSG 1
54 #define TIPC_MAX_PORT 0xffffffff
55 #define TIPC_MIN_PORT 1
56 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
59 TIPC_LISTEN = TCP_LISTEN,
60 TIPC_ESTABLISHED = TCP_ESTABLISHED,
61 TIPC_OPEN = TCP_CLOSE,
62 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
63 TIPC_CONNECTING = TCP_SYN_SENT,
66 struct sockaddr_pair {
67 struct sockaddr_tipc sock;
68 struct sockaddr_tipc member;
72 * struct tipc_sock - TIPC socket structure
73 * @sk: socket - interacts with 'port' and with user via the socket API
74 * @conn_type: TIPC type used when connection was established
75 * @conn_instance: TIPC instance used when connection was established
76 * @published: non-zero if port has one or more associated names
77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
78 * @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 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
86 * @cong_link_cnt: number of congested links
87 * @snt_unacked: # messages sent by socket, and not yet acked by peer
88 * @rcv_unacked: # messages read by user, but not yet acked back to peer
89 * @peer: 'connected' peer for dgram/rdm
90 * @node: hash table node
91 * @mc_method: cookie for use between socket and broadcast layer
92 * @rcu: rcu struct for tipc_sock
101 struct tipc_msg phdr;
102 struct list_head cong_links;
103 struct list_head publications;
105 atomic_t dupl_rcvcnt;
114 struct sockaddr_tipc peer;
115 struct rhash_head node;
116 struct tipc_mc_method mc_method;
118 struct tipc_group *group;
122 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
123 static void tipc_data_ready(struct sock *sk);
124 static void tipc_write_space(struct sock *sk);
125 static void tipc_sock_destruct(struct sock *sk);
126 static int tipc_release(struct socket *sock);
127 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
129 static void tipc_sk_timeout(struct timer_list *t);
130 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
131 struct tipc_name_seq const *seq);
132 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
133 struct tipc_name_seq const *seq);
134 static int tipc_sk_leave(struct tipc_sock *tsk);
135 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
136 static int tipc_sk_insert(struct tipc_sock *tsk);
137 static void tipc_sk_remove(struct tipc_sock *tsk);
138 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
139 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
141 static const struct proto_ops packet_ops;
142 static const struct proto_ops stream_ops;
143 static const struct proto_ops msg_ops;
144 static struct proto tipc_proto;
145 static const struct rhashtable_params tsk_rht_params;
147 static u32 tsk_own_node(struct tipc_sock *tsk)
149 return msg_prevnode(&tsk->phdr);
152 static u32 tsk_peer_node(struct tipc_sock *tsk)
154 return msg_destnode(&tsk->phdr);
157 static u32 tsk_peer_port(struct tipc_sock *tsk)
159 return msg_destport(&tsk->phdr);
162 static bool tsk_unreliable(struct tipc_sock *tsk)
164 return msg_src_droppable(&tsk->phdr) != 0;
167 static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
169 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
172 static bool tsk_unreturnable(struct tipc_sock *tsk)
174 return msg_dest_droppable(&tsk->phdr) != 0;
177 static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
179 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
182 static int tsk_importance(struct tipc_sock *tsk)
184 return msg_importance(&tsk->phdr);
187 static int tsk_set_importance(struct tipc_sock *tsk, int imp)
189 if (imp > TIPC_CRITICAL_IMPORTANCE)
191 msg_set_importance(&tsk->phdr, (u32)imp);
195 static struct tipc_sock *tipc_sk(const struct sock *sk)
197 return container_of(sk, struct tipc_sock, sk);
200 static bool tsk_conn_cong(struct tipc_sock *tsk)
202 return tsk->snt_unacked > tsk->snd_win;
205 static u16 tsk_blocks(int len)
207 return ((len / FLOWCTL_BLK_SZ) + 1);
210 /* tsk_blocks(): translate a buffer size in bytes to number of
211 * advertisable blocks, taking into account the ratio truesize(len)/len
212 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
214 static u16 tsk_adv_blocks(int len)
216 return len / FLOWCTL_BLK_SZ / 4;
219 /* tsk_inc(): increment counter for sent or received data
220 * - If block based flow control is not supported by peer we
221 * fall back to message based ditto, incrementing the counter
223 static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
225 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
226 return ((msglen / FLOWCTL_BLK_SZ) + 1);
231 * tsk_advance_rx_queue - discard first buffer in socket receive queue
233 * Caller must hold socket lock
235 static void tsk_advance_rx_queue(struct sock *sk)
237 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
238 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
241 /* tipc_sk_respond() : send response message back to sender
243 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
247 u32 onode = tipc_own_addr(sock_net(sk));
249 if (!tipc_msg_reverse(onode, &skb, err))
252 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
253 dnode = msg_destnode(buf_msg(skb));
254 selector = msg_origport(buf_msg(skb));
255 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
259 * tsk_rej_rx_queue - reject all buffers in socket receive queue
261 * Caller must hold socket lock
263 static void tsk_rej_rx_queue(struct sock *sk)
267 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
268 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
271 static bool tipc_sk_connected(struct sock *sk)
273 return sk->sk_state == TIPC_ESTABLISHED;
276 /* tipc_sk_type_connectionless - check if the socket is datagram socket
279 * Returns true if connection less, false otherwise
281 static bool tipc_sk_type_connectionless(struct sock *sk)
283 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
286 /* tsk_peer_msg - verify if message was sent by connected port's peer
288 * Handles cases where the node's network address has changed from
289 * the default of <0.0.0> to its configured setting.
291 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
293 struct sock *sk = &tsk->sk;
294 u32 self = tipc_own_addr(sock_net(sk));
295 u32 peer_port = tsk_peer_port(tsk);
296 u32 orig_node, peer_node;
298 if (unlikely(!tipc_sk_connected(sk)))
301 if (unlikely(msg_origport(msg) != peer_port))
304 orig_node = msg_orignode(msg);
305 peer_node = tsk_peer_node(tsk);
307 if (likely(orig_node == peer_node))
310 if (!orig_node && peer_node == self)
313 if (!peer_node && orig_node == self)
319 /* tipc_set_sk_state - set the sk_state of the socket
322 * Caller must hold socket lock
324 * Returns 0 on success, errno otherwise
326 static int tipc_set_sk_state(struct sock *sk, int state)
328 int oldsk_state = sk->sk_state;
336 case TIPC_CONNECTING:
337 if (oldsk_state == TIPC_OPEN)
340 case TIPC_ESTABLISHED:
341 if (oldsk_state == TIPC_CONNECTING ||
342 oldsk_state == TIPC_OPEN)
345 case TIPC_DISCONNECTING:
346 if (oldsk_state == TIPC_CONNECTING ||
347 oldsk_state == TIPC_ESTABLISHED)
353 sk->sk_state = state;
358 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
360 struct sock *sk = sock->sk;
361 int err = sock_error(sk);
362 int typ = sock->type;
366 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
367 if (sk->sk_state == TIPC_DISCONNECTING)
369 else if (!tipc_sk_connected(sk))
374 if (signal_pending(current))
375 return sock_intr_errno(*timeout);
380 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
382 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
386 while ((rc_ = !(condition_))) { \
387 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
390 rc_ = tipc_sk_sock_err((sock_), timeo_); \
393 add_wait_queue(sk_sleep(sk_), &wait_); \
395 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
396 sched_annotate_sleep(); \
398 remove_wait_queue(sk_sleep(sk_), &wait_); \
404 * tipc_sk_create - create a TIPC socket
405 * @net: network namespace (must be default network)
406 * @sock: pre-allocated socket structure
407 * @protocol: protocol indicator (must be 0)
408 * @kern: caused by kernel or by userspace?
410 * This routine creates additional data structures used by the TIPC socket,
411 * initializes them, and links them together.
413 * Returns 0 on success, errno otherwise
415 static int tipc_sk_create(struct net *net, struct socket *sock,
416 int protocol, int kern)
418 const struct proto_ops *ops;
420 struct tipc_sock *tsk;
421 struct tipc_msg *msg;
423 /* Validate arguments */
424 if (unlikely(protocol != 0))
425 return -EPROTONOSUPPORT;
427 switch (sock->type) {
442 /* Allocate socket's protocol area */
443 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
448 tsk->max_pkt = MAX_PKT_DEFAULT;
449 INIT_LIST_HEAD(&tsk->publications);
450 INIT_LIST_HEAD(&tsk->cong_links);
453 /* Finish initializing socket data structures */
455 sock_init_data(sock, sk);
456 tipc_set_sk_state(sk, TIPC_OPEN);
457 if (tipc_sk_insert(tsk)) {
458 pr_warn("Socket create failed; port number exhausted\n");
462 /* Ensure tsk is visible before we read own_addr. */
465 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
466 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
468 msg_set_origport(msg, tsk->portid);
469 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
471 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
472 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
473 sk->sk_data_ready = tipc_data_ready;
474 sk->sk_write_space = tipc_write_space;
475 sk->sk_destruct = tipc_sock_destruct;
476 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
477 tsk->group_is_open = true;
478 atomic_set(&tsk->dupl_rcvcnt, 0);
480 /* Start out with safe limits until we receive an advertised window */
481 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
482 tsk->rcv_win = tsk->snd_win;
484 if (tipc_sk_type_connectionless(sk)) {
485 tsk_set_unreturnable(tsk, true);
486 if (sock->type == SOCK_DGRAM)
487 tsk_set_unreliable(tsk, true);
490 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
494 static void tipc_sk_callback(struct rcu_head *head)
496 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
501 /* Caller should hold socket lock for the socket. */
502 static void __tipc_shutdown(struct socket *sock, int error)
504 struct sock *sk = sock->sk;
505 struct tipc_sock *tsk = tipc_sk(sk);
506 struct net *net = sock_net(sk);
507 long timeout = CONN_TIMEOUT_DEFAULT;
508 u32 dnode = tsk_peer_node(tsk);
511 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
512 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
513 !tsk_conn_cong(tsk)));
515 /* Remove any pending SYN message */
516 __skb_queue_purge(&sk->sk_write_queue);
518 /* Reject all unreceived messages, except on an active connection
519 * (which disconnects locally & sends a 'FIN+' to peer).
521 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
522 if (TIPC_SKB_CB(skb)->bytes_read) {
526 if (!tipc_sk_type_connectionless(sk) &&
527 sk->sk_state != TIPC_DISCONNECTING) {
528 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
529 tipc_node_remove_conn(net, dnode, tsk->portid);
531 tipc_sk_respond(sk, skb, error);
534 if (tipc_sk_type_connectionless(sk))
537 if (sk->sk_state != TIPC_DISCONNECTING) {
538 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
539 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
540 tsk_own_node(tsk), tsk_peer_port(tsk),
543 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
544 tipc_node_remove_conn(net, dnode, tsk->portid);
545 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
550 * tipc_release - destroy a TIPC socket
551 * @sock: socket to destroy
553 * This routine cleans up any messages that are still queued on the socket.
554 * For DGRAM and RDM socket types, all queued messages are rejected.
555 * For SEQPACKET and STREAM socket types, the first message is rejected
556 * and any others are discarded. (If the first message on a STREAM socket
557 * is partially-read, it is discarded and the next one is rejected instead.)
559 * NOTE: Rejected messages are not necessarily returned to the sender! They
560 * are returned or discarded according to the "destination droppable" setting
561 * specified for the message by the sender.
563 * Returns 0 on success, errno otherwise
565 static int tipc_release(struct socket *sock)
567 struct sock *sk = sock->sk;
568 struct tipc_sock *tsk;
571 * Exit if socket isn't fully initialized (occurs when a failed accept()
572 * releases a pre-allocated child socket that was never used)
580 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
581 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
582 sk->sk_shutdown = SHUTDOWN_MASK;
584 tipc_sk_withdraw(tsk, 0, NULL);
585 sk_stop_timer(sk, &sk->sk_timer);
589 /* Reject any messages that accumulated in backlog queue */
591 tipc_dest_list_purge(&tsk->cong_links);
592 tsk->cong_link_cnt = 0;
593 call_rcu(&tsk->rcu, tipc_sk_callback);
600 * tipc_bind - associate or disassocate TIPC name(s) with a socket
601 * @sock: socket structure
602 * @uaddr: socket address describing name(s) and desired operation
603 * @uaddr_len: size of socket address data structure
605 * Name and name sequence binding is indicated using a positive scope value;
606 * a negative scope value unbinds the specified name. Specifying no name
607 * (i.e. a socket address length of 0) unbinds all names from the socket.
609 * Returns 0 on success, errno otherwise
611 * NOTE: This routine doesn't need to take the socket lock since it doesn't
612 * access any non-constant socket information.
614 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
617 struct sock *sk = sock->sk;
618 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
619 struct tipc_sock *tsk = tipc_sk(sk);
623 if (unlikely(!uaddr_len)) {
624 res = tipc_sk_withdraw(tsk, 0, NULL);
631 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
635 if (addr->family != AF_TIPC) {
640 if (addr->addrtype == TIPC_ADDR_NAME)
641 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
642 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
647 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
648 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
649 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
654 res = (addr->scope >= 0) ?
655 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
656 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
663 * tipc_getname - get port ID of socket or peer socket
664 * @sock: socket structure
665 * @uaddr: area for returned socket address
666 * @uaddr_len: area for returned length of socket address
667 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
669 * Returns 0 on success, errno otherwise
671 * NOTE: This routine doesn't need to take the socket lock since it only
672 * accesses socket information that is unchanging (or which changes in
673 * a completely predictable manner).
675 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
678 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
679 struct sock *sk = sock->sk;
680 struct tipc_sock *tsk = tipc_sk(sk);
682 memset(addr, 0, sizeof(*addr));
684 if ((!tipc_sk_connected(sk)) &&
685 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
687 addr->addr.id.ref = tsk_peer_port(tsk);
688 addr->addr.id.node = tsk_peer_node(tsk);
690 addr->addr.id.ref = tsk->portid;
691 addr->addr.id.node = tipc_own_addr(sock_net(sk));
694 addr->addrtype = TIPC_ADDR_ID;
695 addr->family = AF_TIPC;
697 addr->addr.name.domain = 0;
699 return sizeof(*addr);
703 * tipc_poll - read and possibly block on pollmask
704 * @file: file structure associated with the socket
705 * @sock: socket for which to calculate the poll bits
708 * Returns pollmask value
711 * It appears that the usual socket locking mechanisms are not useful here
712 * since the pollmask info is potentially out-of-date the moment this routine
713 * exits. TCP and other protocols seem to rely on higher level poll routines
714 * to handle any preventable race conditions, so TIPC will do the same ...
716 * IMPORTANT: The fact that a read or write operation is indicated does NOT
717 * imply that the operation will succeed, merely that it should be performed
718 * and will not block.
720 static __poll_t tipc_poll(struct file *file, struct socket *sock,
723 struct sock *sk = sock->sk;
724 struct tipc_sock *tsk = tipc_sk(sk);
725 __poll_t revents = 0;
727 sock_poll_wait(file, sock, wait);
728 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
730 if (sk->sk_shutdown & RCV_SHUTDOWN)
731 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
732 if (sk->sk_shutdown == SHUTDOWN_MASK)
735 switch (sk->sk_state) {
736 case TIPC_ESTABLISHED:
737 case TIPC_CONNECTING:
738 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
742 if (!skb_queue_empty(&sk->sk_receive_queue))
743 revents |= EPOLLIN | EPOLLRDNORM;
746 if (tsk->group_is_open && !tsk->cong_link_cnt)
748 if (!tipc_sk_type_connectionless(sk))
750 if (skb_queue_empty(&sk->sk_receive_queue))
752 revents |= EPOLLIN | EPOLLRDNORM;
754 case TIPC_DISCONNECTING:
755 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
762 * tipc_sendmcast - send multicast message
763 * @sock: socket structure
764 * @seq: destination address
765 * @msg: message to send
766 * @dlen: length of data to send
767 * @timeout: timeout to wait for wakeup
769 * Called from function tipc_sendmsg(), which has done all sanity checks
770 * Returns the number of bytes sent on success, or errno
772 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
773 struct msghdr *msg, size_t dlen, long timeout)
775 struct sock *sk = sock->sk;
776 struct tipc_sock *tsk = tipc_sk(sk);
777 struct tipc_msg *hdr = &tsk->phdr;
778 struct net *net = sock_net(sk);
779 int mtu = tipc_bcast_get_mtu(net);
780 struct tipc_mc_method *method = &tsk->mc_method;
781 struct sk_buff_head pkts;
782 struct tipc_nlist dsts;
788 /* Block or return if any destination link is congested */
789 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
793 /* Lookup destination nodes */
794 tipc_nlist_init(&dsts, tipc_own_addr(net));
795 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
797 if (!dsts.local && !dsts.remote)
798 return -EHOSTUNREACH;
800 /* Build message header */
801 msg_set_type(hdr, TIPC_MCAST_MSG);
802 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
803 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
804 msg_set_destport(hdr, 0);
805 msg_set_destnode(hdr, 0);
806 msg_set_nametype(hdr, seq->type);
807 msg_set_namelower(hdr, seq->lower);
808 msg_set_nameupper(hdr, seq->upper);
810 /* Build message as chain of buffers */
811 skb_queue_head_init(&pkts);
812 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
814 /* Send message if build was successful */
815 if (unlikely(rc == dlen)) {
816 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
817 TIPC_DUMP_SK_SNDQ, " ");
818 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
819 &tsk->cong_link_cnt);
822 tipc_nlist_purge(&dsts);
824 return rc ? rc : dlen;
828 * tipc_send_group_msg - send a message to a member in the group
829 * @net: network namespace
830 * @m: message to send
832 * @dnode: destination node
833 * @dport: destination port
834 * @dlen: total length of message data
836 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
837 struct msghdr *m, struct tipc_member *mb,
838 u32 dnode, u32 dport, int dlen)
840 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
841 struct tipc_mc_method *method = &tsk->mc_method;
842 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
843 struct tipc_msg *hdr = &tsk->phdr;
844 struct sk_buff_head pkts;
847 /* Complete message header */
848 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
849 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
850 msg_set_destport(hdr, dport);
851 msg_set_destnode(hdr, dnode);
852 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
854 /* Build message as chain of buffers */
855 skb_queue_head_init(&pkts);
856 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
857 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
858 if (unlikely(rc != dlen))
862 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
863 if (unlikely(rc == -ELINKCONG)) {
864 tipc_dest_push(&tsk->cong_links, dnode, 0);
865 tsk->cong_link_cnt++;
868 /* Update send window */
869 tipc_group_update_member(mb, blks);
871 /* A broadcast sent within next EXPIRE period must follow same path */
872 method->rcast = true;
873 method->mandatory = true;
878 * tipc_send_group_unicast - send message to a member in the group
879 * @sock: socket structure
880 * @m: message to send
881 * @dlen: total length of message data
882 * @timeout: timeout to wait for wakeup
884 * Called from function tipc_sendmsg(), which has done all sanity checks
885 * Returns the number of bytes sent on success, or errno
887 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
888 int dlen, long timeout)
890 struct sock *sk = sock->sk;
891 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
892 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
893 struct tipc_sock *tsk = tipc_sk(sk);
894 struct net *net = sock_net(sk);
895 struct tipc_member *mb = NULL;
899 node = dest->addr.id.node;
900 port = dest->addr.id.ref;
902 return -EHOSTUNREACH;
904 /* Block or return if destination link or member is congested */
905 rc = tipc_wait_for_cond(sock, &timeout,
906 !tipc_dest_find(&tsk->cong_links, node, 0) &&
908 !tipc_group_cong(tsk->group, node, port, blks,
914 return -EHOSTUNREACH;
916 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
918 return rc ? rc : dlen;
922 * tipc_send_group_anycast - send message to any member with given identity
923 * @sock: socket structure
924 * @m: message to send
925 * @dlen: total length of message data
926 * @timeout: timeout to wait for wakeup
928 * Called from function tipc_sendmsg(), which has done all sanity checks
929 * Returns the number of bytes sent on success, or errno
931 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
932 int dlen, long timeout)
934 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
935 struct sock *sk = sock->sk;
936 struct tipc_sock *tsk = tipc_sk(sk);
937 struct list_head *cong_links = &tsk->cong_links;
938 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
939 struct tipc_msg *hdr = &tsk->phdr;
940 struct tipc_member *first = NULL;
941 struct tipc_member *mbr = NULL;
942 struct net *net = sock_net(sk);
943 u32 node, port, exclude;
944 struct list_head dsts;
945 u32 type, inst, scope;
950 INIT_LIST_HEAD(&dsts);
952 type = msg_nametype(hdr);
953 inst = dest->addr.name.name.instance;
954 scope = msg_lookup_scope(hdr);
956 while (++lookups < 4) {
957 exclude = tipc_group_exclude(tsk->group);
961 /* Look for a non-congested destination member, if any */
963 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
964 &dstcnt, exclude, false))
965 return -EHOSTUNREACH;
966 tipc_dest_pop(&dsts, &node, &port);
967 cong = tipc_group_cong(tsk->group, node, port, blks,
977 /* Start over if destination was not in member list */
981 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
984 /* Block or return if destination link or member is congested */
985 rc = tipc_wait_for_cond(sock, &timeout,
986 !tipc_dest_find(cong_links, node, 0) &&
988 !tipc_group_cong(tsk->group, node, port,
993 /* Send, unless destination disappeared while waiting */
998 if (unlikely(lookups >= 4))
999 return -EHOSTUNREACH;
1001 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1003 return rc ? rc : dlen;
1007 * tipc_send_group_bcast - send message to all members in communication group
1008 * @sk: socket structure
1009 * @m: message to send
1010 * @dlen: total length of message data
1011 * @timeout: timeout to wait for wakeup
1013 * Called from function tipc_sendmsg(), which has done all sanity checks
1014 * Returns the number of bytes sent on success, or errno
1016 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1017 int dlen, long timeout)
1019 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1020 struct sock *sk = sock->sk;
1021 struct net *net = sock_net(sk);
1022 struct tipc_sock *tsk = tipc_sk(sk);
1023 struct tipc_nlist *dsts;
1024 struct tipc_mc_method *method = &tsk->mc_method;
1025 bool ack = method->mandatory && method->rcast;
1026 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1027 struct tipc_msg *hdr = &tsk->phdr;
1028 int mtu = tipc_bcast_get_mtu(net);
1029 struct sk_buff_head pkts;
1030 int rc = -EHOSTUNREACH;
1032 /* Block or return if any destination link or member is congested */
1033 rc = tipc_wait_for_cond(sock, &timeout,
1034 !tsk->cong_link_cnt && tsk->group &&
1035 !tipc_group_bc_cong(tsk->group, blks));
1039 dsts = tipc_group_dests(tsk->group);
1040 if (!dsts->local && !dsts->remote)
1041 return -EHOSTUNREACH;
1043 /* Complete message header */
1045 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1046 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1048 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1049 msg_set_nameinst(hdr, 0);
1051 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1052 msg_set_destport(hdr, 0);
1053 msg_set_destnode(hdr, 0);
1054 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1056 /* Avoid getting stuck with repeated forced replicasts */
1057 msg_set_grp_bc_ack_req(hdr, ack);
1059 /* Build message as chain of buffers */
1060 skb_queue_head_init(&pkts);
1061 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1062 if (unlikely(rc != dlen))
1066 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1070 /* Update broadcast sequence number and send windows */
1071 tipc_group_update_bc_members(tsk->group, blks, ack);
1073 /* Broadcast link is now free to choose method for next broadcast */
1074 method->mandatory = false;
1075 method->expires = jiffies;
1081 * tipc_send_group_mcast - send message to all members with given identity
1082 * @sock: socket structure
1083 * @m: message to send
1084 * @dlen: total length of message data
1085 * @timeout: timeout to wait for wakeup
1087 * Called from function tipc_sendmsg(), which has done all sanity checks
1088 * Returns the number of bytes sent on success, or errno
1090 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1091 int dlen, long timeout)
1093 struct sock *sk = sock->sk;
1094 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1095 struct tipc_sock *tsk = tipc_sk(sk);
1096 struct tipc_group *grp = tsk->group;
1097 struct tipc_msg *hdr = &tsk->phdr;
1098 struct net *net = sock_net(sk);
1099 u32 type, inst, scope, exclude;
1100 struct list_head dsts;
1103 INIT_LIST_HEAD(&dsts);
1105 type = msg_nametype(hdr);
1106 inst = dest->addr.name.name.instance;
1107 scope = msg_lookup_scope(hdr);
1108 exclude = tipc_group_exclude(grp);
1110 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1111 &dstcnt, exclude, true))
1112 return -EHOSTUNREACH;
1115 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1116 return tipc_send_group_unicast(sock, m, dlen, timeout);
1119 tipc_dest_list_purge(&dsts);
1120 return tipc_send_group_bcast(sock, m, dlen, timeout);
1124 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1125 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1126 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1128 * Multi-threaded: parallel calls with reference to same queues may occur
1130 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1131 struct sk_buff_head *inputq)
1133 u32 self = tipc_own_addr(net);
1134 u32 type, lower, upper, scope;
1135 struct sk_buff *skb, *_skb;
1137 struct sk_buff_head tmpq;
1138 struct list_head dports;
1139 struct tipc_msg *hdr;
1140 int user, mtyp, hlen;
1143 __skb_queue_head_init(&tmpq);
1144 INIT_LIST_HEAD(&dports);
1146 skb = tipc_skb_peek(arrvq, &inputq->lock);
1147 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1149 user = msg_user(hdr);
1150 mtyp = msg_type(hdr);
1151 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1152 onode = msg_orignode(hdr);
1153 type = msg_nametype(hdr);
1155 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1156 spin_lock_bh(&inputq->lock);
1157 if (skb_peek(arrvq) == skb) {
1158 __skb_dequeue(arrvq);
1159 __skb_queue_tail(inputq, skb);
1162 spin_unlock_bh(&inputq->lock);
1166 /* Group messages require exact scope match */
1167 if (msg_in_group(hdr)) {
1170 scope = msg_lookup_scope(hdr);
1173 /* TIPC_NODE_SCOPE means "any scope" in this context */
1175 scope = TIPC_NODE_SCOPE;
1177 scope = TIPC_CLUSTER_SCOPE;
1179 lower = msg_namelower(hdr);
1180 upper = msg_nameupper(hdr);
1183 /* Create destination port list: */
1184 tipc_nametbl_mc_lookup(net, type, lower, upper,
1185 scope, exact, &dports);
1187 /* Clone message per destination */
1188 while (tipc_dest_pop(&dports, NULL, &portid)) {
1189 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1191 msg_set_destport(buf_msg(_skb), portid);
1192 __skb_queue_tail(&tmpq, _skb);
1195 pr_warn("Failed to clone mcast rcv buffer\n");
1197 /* Append to inputq if not already done by other thread */
1198 spin_lock_bh(&inputq->lock);
1199 if (skb_peek(arrvq) == skb) {
1200 skb_queue_splice_tail_init(&tmpq, inputq);
1201 kfree_skb(__skb_dequeue(arrvq));
1203 spin_unlock_bh(&inputq->lock);
1204 __skb_queue_purge(&tmpq);
1207 tipc_sk_rcv(net, inputq);
1211 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1212 * @tsk: receiving socket
1213 * @skb: pointer to message buffer.
1215 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1216 struct sk_buff_head *inputq,
1217 struct sk_buff_head *xmitq)
1219 struct tipc_msg *hdr = buf_msg(skb);
1220 u32 onode = tsk_own_node(tsk);
1221 struct sock *sk = &tsk->sk;
1222 int mtyp = msg_type(hdr);
1225 /* Ignore if connection cannot be validated: */
1226 if (!tsk_peer_msg(tsk, hdr)) {
1227 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1231 if (unlikely(msg_errcode(hdr))) {
1232 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1233 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1234 tsk_peer_port(tsk));
1235 sk->sk_state_change(sk);
1237 /* State change is ignored if socket already awake,
1238 * - convert msg to abort msg and add to inqueue
1240 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1241 msg_set_type(hdr, TIPC_CONN_MSG);
1242 msg_set_size(hdr, BASIC_H_SIZE);
1243 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1244 __skb_queue_tail(inputq, skb);
1248 tsk->probe_unacked = false;
1250 if (mtyp == CONN_PROBE) {
1251 msg_set_type(hdr, CONN_PROBE_REPLY);
1252 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1253 __skb_queue_tail(xmitq, skb);
1255 } else if (mtyp == CONN_ACK) {
1256 conn_cong = tsk_conn_cong(tsk);
1257 tsk->snt_unacked -= msg_conn_ack(hdr);
1258 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1259 tsk->snd_win = msg_adv_win(hdr);
1261 sk->sk_write_space(sk);
1262 } else if (mtyp != CONN_PROBE_REPLY) {
1263 pr_warn("Received unknown CONN_PROTO msg\n");
1270 * tipc_sendmsg - send message in connectionless manner
1271 * @sock: socket structure
1272 * @m: message to send
1273 * @dsz: amount of user data to be sent
1275 * Message must have an destination specified explicitly.
1276 * Used for SOCK_RDM and SOCK_DGRAM messages,
1277 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1278 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1280 * Returns the number of bytes sent on success, or errno otherwise
1282 static int tipc_sendmsg(struct socket *sock,
1283 struct msghdr *m, size_t dsz)
1285 struct sock *sk = sock->sk;
1289 ret = __tipc_sendmsg(sock, m, dsz);
1295 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1297 struct sock *sk = sock->sk;
1298 struct net *net = sock_net(sk);
1299 struct tipc_sock *tsk = tipc_sk(sk);
1300 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1301 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1302 struct list_head *clinks = &tsk->cong_links;
1303 bool syn = !tipc_sk_type_connectionless(sk);
1304 struct tipc_group *grp = tsk->group;
1305 struct tipc_msg *hdr = &tsk->phdr;
1306 struct tipc_name_seq *seq;
1307 struct sk_buff_head pkts;
1308 u32 dport, dnode = 0;
1312 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1316 if (unlikely(m->msg_namelen < sizeof(*dest)))
1318 if (unlikely(dest->family != AF_TIPC))
1324 return tipc_send_group_bcast(sock, m, dlen, timeout);
1325 if (dest->addrtype == TIPC_ADDR_NAME)
1326 return tipc_send_group_anycast(sock, m, dlen, timeout);
1327 if (dest->addrtype == TIPC_ADDR_ID)
1328 return tipc_send_group_unicast(sock, m, dlen, timeout);
1329 if (dest->addrtype == TIPC_ADDR_MCAST)
1330 return tipc_send_group_mcast(sock, m, dlen, timeout);
1334 if (unlikely(!dest)) {
1336 if (!syn && dest->family != AF_TIPC)
1337 return -EDESTADDRREQ;
1340 if (unlikely(syn)) {
1341 if (sk->sk_state == TIPC_LISTEN)
1343 if (sk->sk_state != TIPC_OPEN)
1347 if (dest->addrtype == TIPC_ADDR_NAME) {
1348 tsk->conn_type = dest->addr.name.name.type;
1349 tsk->conn_instance = dest->addr.name.name.instance;
1351 msg_set_syn(hdr, 1);
1354 seq = &dest->addr.nameseq;
1355 if (dest->addrtype == TIPC_ADDR_MCAST)
1356 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1358 if (dest->addrtype == TIPC_ADDR_NAME) {
1359 type = dest->addr.name.name.type;
1360 inst = dest->addr.name.name.instance;
1361 dnode = dest->addr.name.domain;
1362 msg_set_type(hdr, TIPC_NAMED_MSG);
1363 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1364 msg_set_nametype(hdr, type);
1365 msg_set_nameinst(hdr, inst);
1366 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1367 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1368 msg_set_destnode(hdr, dnode);
1369 msg_set_destport(hdr, dport);
1370 if (unlikely(!dport && !dnode))
1371 return -EHOSTUNREACH;
1372 } else if (dest->addrtype == TIPC_ADDR_ID) {
1373 dnode = dest->addr.id.node;
1374 msg_set_type(hdr, TIPC_DIRECT_MSG);
1375 msg_set_lookup_scope(hdr, 0);
1376 msg_set_destnode(hdr, dnode);
1377 msg_set_destport(hdr, dest->addr.id.ref);
1378 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1383 /* Block or return if destination link is congested */
1384 rc = tipc_wait_for_cond(sock, &timeout,
1385 !tipc_dest_find(clinks, dnode, 0));
1389 skb_queue_head_init(&pkts);
1390 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1391 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1392 if (unlikely(rc != dlen))
1394 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue)))
1397 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1398 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1399 if (unlikely(rc == -ELINKCONG)) {
1400 tipc_dest_push(clinks, dnode, 0);
1401 tsk->cong_link_cnt++;
1405 if (unlikely(syn && !rc))
1406 tipc_set_sk_state(sk, TIPC_CONNECTING);
1408 return rc ? rc : dlen;
1412 * tipc_sendstream - send stream-oriented data
1413 * @sock: socket structure
1415 * @dsz: total length of data to be transmitted
1417 * Used for SOCK_STREAM data.
1419 * Returns the number of bytes sent on success (or partial success),
1420 * or errno if no data sent
1422 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1424 struct sock *sk = sock->sk;
1428 ret = __tipc_sendstream(sock, m, dsz);
1434 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1436 struct sock *sk = sock->sk;
1437 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1438 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1439 struct tipc_sock *tsk = tipc_sk(sk);
1440 struct tipc_msg *hdr = &tsk->phdr;
1441 struct net *net = sock_net(sk);
1442 struct sk_buff_head pkts;
1443 u32 dnode = tsk_peer_node(tsk);
1447 skb_queue_head_init(&pkts);
1449 if (unlikely(dlen > INT_MAX))
1452 /* Handle implicit connection setup */
1453 if (unlikely(dest)) {
1454 rc = __tipc_sendmsg(sock, m, dlen);
1455 if (dlen && dlen == rc) {
1456 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1457 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1463 rc = tipc_wait_for_cond(sock, &timeout,
1464 (!tsk->cong_link_cnt &&
1465 !tsk_conn_cong(tsk) &&
1466 tipc_sk_connected(sk)));
1470 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1471 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1472 if (unlikely(rc != send))
1475 trace_tipc_sk_sendstream(sk, skb_peek(&pkts),
1476 TIPC_DUMP_SK_SNDQ, " ");
1477 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1478 if (unlikely(rc == -ELINKCONG)) {
1479 tsk->cong_link_cnt = 1;
1483 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1486 } while (sent < dlen && !rc);
1488 return sent ? sent : rc;
1492 * tipc_send_packet - send a connection-oriented message
1493 * @sock: socket structure
1494 * @m: message to send
1495 * @dsz: length of data to be transmitted
1497 * Used for SOCK_SEQPACKET messages.
1499 * Returns the number of bytes sent on success, or errno otherwise
1501 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1503 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1506 return tipc_sendstream(sock, m, dsz);
1509 /* tipc_sk_finish_conn - complete the setup of a connection
1511 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1514 struct sock *sk = &tsk->sk;
1515 struct net *net = sock_net(sk);
1516 struct tipc_msg *msg = &tsk->phdr;
1518 msg_set_syn(msg, 0);
1519 msg_set_destnode(msg, peer_node);
1520 msg_set_destport(msg, peer_port);
1521 msg_set_type(msg, TIPC_CONN_MSG);
1522 msg_set_lookup_scope(msg, 0);
1523 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1525 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1526 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1527 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1528 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1529 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1530 __skb_queue_purge(&sk->sk_write_queue);
1531 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1534 /* Fall back to message based flow control */
1535 tsk->rcv_win = FLOWCTL_MSG_WIN;
1536 tsk->snd_win = FLOWCTL_MSG_WIN;
1540 * tipc_sk_set_orig_addr - capture sender's address for received message
1541 * @m: descriptor for message info
1542 * @hdr: received message header
1544 * Note: Address is not captured if not requested by receiver.
1546 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1548 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1549 struct tipc_msg *hdr = buf_msg(skb);
1554 srcaddr->sock.family = AF_TIPC;
1555 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1556 srcaddr->sock.scope = 0;
1557 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1558 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1559 srcaddr->sock.addr.name.domain = 0;
1560 m->msg_namelen = sizeof(struct sockaddr_tipc);
1562 if (!msg_in_group(hdr))
1565 /* Group message users may also want to know sending member's id */
1566 srcaddr->member.family = AF_TIPC;
1567 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1568 srcaddr->member.scope = 0;
1569 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1570 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1571 srcaddr->member.addr.name.domain = 0;
1572 m->msg_namelen = sizeof(*srcaddr);
1576 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1577 * @m: descriptor for message info
1578 * @skb: received message buffer
1579 * @tsk: TIPC port associated with message
1581 * Note: Ancillary data is not captured if not requested by receiver.
1583 * Returns 0 if successful, otherwise errno
1585 static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1586 struct tipc_sock *tsk)
1588 struct tipc_msg *msg;
1595 if (likely(m->msg_controllen == 0))
1599 /* Optionally capture errored message object(s) */
1600 err = msg ? msg_errcode(msg) : 0;
1601 if (unlikely(err)) {
1603 anc_data[1] = msg_data_sz(msg);
1604 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1608 if (skb_linearize(skb))
1611 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1618 /* Optionally capture message destination object */
1619 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1620 switch (dest_type) {
1621 case TIPC_NAMED_MSG:
1623 anc_data[0] = msg_nametype(msg);
1624 anc_data[1] = msg_namelower(msg);
1625 anc_data[2] = msg_namelower(msg);
1627 case TIPC_MCAST_MSG:
1629 anc_data[0] = msg_nametype(msg);
1630 anc_data[1] = msg_namelower(msg);
1631 anc_data[2] = msg_nameupper(msg);
1634 has_name = (tsk->conn_type != 0);
1635 anc_data[0] = tsk->conn_type;
1636 anc_data[1] = tsk->conn_instance;
1637 anc_data[2] = tsk->conn_instance;
1643 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1651 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1653 struct sock *sk = &tsk->sk;
1654 struct net *net = sock_net(sk);
1655 struct sk_buff *skb = NULL;
1656 struct tipc_msg *msg;
1657 u32 peer_port = tsk_peer_port(tsk);
1658 u32 dnode = tsk_peer_node(tsk);
1660 if (!tipc_sk_connected(sk))
1662 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1663 dnode, tsk_own_node(tsk), peer_port,
1664 tsk->portid, TIPC_OK);
1668 msg_set_conn_ack(msg, tsk->rcv_unacked);
1669 tsk->rcv_unacked = 0;
1671 /* Adjust to and advertize the correct window limit */
1672 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1673 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1674 msg_set_adv_win(msg, tsk->rcv_win);
1676 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1679 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1681 struct sock *sk = sock->sk;
1682 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1683 long timeo = *timeop;
1684 int err = sock_error(sk);
1690 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1691 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1695 add_wait_queue(sk_sleep(sk), &wait);
1697 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1698 sched_annotate_sleep();
1700 remove_wait_queue(sk_sleep(sk), &wait);
1703 if (!skb_queue_empty(&sk->sk_receive_queue))
1708 err = sock_intr_errno(timeo);
1709 if (signal_pending(current))
1712 err = sock_error(sk);
1721 * tipc_recvmsg - receive packet-oriented message
1722 * @m: descriptor for message info
1723 * @buflen: length of user buffer area
1724 * @flags: receive flags
1726 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1727 * If the complete message doesn't fit in user area, truncate it.
1729 * Returns size of returned message data, errno otherwise
1731 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1732 size_t buflen, int flags)
1734 struct sock *sk = sock->sk;
1735 bool connected = !tipc_sk_type_connectionless(sk);
1736 struct tipc_sock *tsk = tipc_sk(sk);
1737 int rc, err, hlen, dlen, copy;
1738 struct sk_buff_head xmitq;
1739 struct tipc_msg *hdr;
1740 struct sk_buff *skb;
1744 /* Catch invalid receive requests */
1745 if (unlikely(!buflen))
1749 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1753 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1755 /* Step rcv queue to first msg with data or error; wait if necessary */
1757 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1760 skb = skb_peek(&sk->sk_receive_queue);
1762 dlen = msg_data_sz(hdr);
1763 hlen = msg_hdr_sz(hdr);
1764 err = msg_errcode(hdr);
1765 grp_evt = msg_is_grp_evt(hdr);
1766 if (likely(dlen || err))
1768 tsk_advance_rx_queue(sk);
1771 /* Collect msg meta data, including error code and rejected data */
1772 tipc_sk_set_orig_addr(m, skb);
1773 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1778 /* Capture data if non-error msg, otherwise just set return value */
1780 copy = min_t(int, dlen, buflen);
1781 if (unlikely(copy != dlen))
1782 m->msg_flags |= MSG_TRUNC;
1783 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1787 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1793 /* Mark message as group event if applicable */
1794 if (unlikely(grp_evt)) {
1795 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1796 m->msg_flags |= MSG_EOR;
1797 m->msg_flags |= MSG_OOB;
1801 /* Caption of data or error code/rejected data was successful */
1802 if (unlikely(flags & MSG_PEEK))
1805 /* Send group flow control advertisement when applicable */
1806 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1807 skb_queue_head_init(&xmitq);
1808 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1809 msg_orignode(hdr), msg_origport(hdr),
1811 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1814 tsk_advance_rx_queue(sk);
1816 if (likely(!connected))
1819 /* Send connection flow control advertisement when applicable */
1820 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1821 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1822 tipc_sk_send_ack(tsk);
1825 return rc ? rc : copy;
1829 * tipc_recvstream - receive stream-oriented data
1830 * @m: descriptor for message info
1831 * @buflen: total size of user buffer area
1832 * @flags: receive flags
1834 * Used for SOCK_STREAM messages only. If not enough data is available
1835 * will optionally wait for more; never truncates data.
1837 * Returns size of returned message data, errno otherwise
1839 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1840 size_t buflen, int flags)
1842 struct sock *sk = sock->sk;
1843 struct tipc_sock *tsk = tipc_sk(sk);
1844 struct sk_buff *skb;
1845 struct tipc_msg *hdr;
1846 struct tipc_skb_cb *skb_cb;
1847 bool peek = flags & MSG_PEEK;
1848 int offset, required, copy, copied = 0;
1849 int hlen, dlen, err, rc;
1852 /* Catch invalid receive attempts */
1853 if (unlikely(!buflen))
1858 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1862 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1863 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1866 /* Look at first msg in receive queue; wait if necessary */
1867 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1870 skb = skb_peek(&sk->sk_receive_queue);
1871 skb_cb = TIPC_SKB_CB(skb);
1873 dlen = msg_data_sz(hdr);
1874 hlen = msg_hdr_sz(hdr);
1875 err = msg_errcode(hdr);
1877 /* Discard any empty non-errored (SYN-) message */
1878 if (unlikely(!dlen && !err)) {
1879 tsk_advance_rx_queue(sk);
1883 /* Collect msg meta data, incl. error code and rejected data */
1885 tipc_sk_set_orig_addr(m, skb);
1886 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1892 /* Copy data if msg ok, otherwise return error/partial data */
1894 offset = skb_cb->bytes_read;
1895 copy = min_t(int, dlen - offset, buflen - copied);
1896 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1901 if (unlikely(offset < dlen)) {
1903 skb_cb->bytes_read = offset;
1908 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1917 tsk_advance_rx_queue(sk);
1919 /* Send connection flow control advertisement when applicable */
1920 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1921 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1922 tipc_sk_send_ack(tsk);
1924 /* Exit if all requested data or FIN/error received */
1925 if (copied == buflen || err)
1928 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1931 return copied ? copied : rc;
1935 * tipc_write_space - wake up thread if port congestion is released
1938 static void tipc_write_space(struct sock *sk)
1940 struct socket_wq *wq;
1943 wq = rcu_dereference(sk->sk_wq);
1944 if (skwq_has_sleeper(wq))
1945 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1946 EPOLLWRNORM | EPOLLWRBAND);
1951 * tipc_data_ready - wake up threads to indicate messages have been received
1953 * @len: the length of messages
1955 static void tipc_data_ready(struct sock *sk)
1957 struct socket_wq *wq;
1960 wq = rcu_dereference(sk->sk_wq);
1961 if (skwq_has_sleeper(wq))
1962 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1963 EPOLLRDNORM | EPOLLRDBAND);
1967 static void tipc_sock_destruct(struct sock *sk)
1969 __skb_queue_purge(&sk->sk_receive_queue);
1972 static void tipc_sk_proto_rcv(struct sock *sk,
1973 struct sk_buff_head *inputq,
1974 struct sk_buff_head *xmitq)
1976 struct sk_buff *skb = __skb_dequeue(inputq);
1977 struct tipc_sock *tsk = tipc_sk(sk);
1978 struct tipc_msg *hdr = buf_msg(skb);
1979 struct tipc_group *grp = tsk->group;
1980 bool wakeup = false;
1982 switch (msg_user(hdr)) {
1984 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
1987 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1988 /* coupled with smp_rmb() in tipc_wait_for_cond() */
1990 tsk->cong_link_cnt--;
1993 case GROUP_PROTOCOL:
1994 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1997 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1998 hdr, inputq, xmitq);
2005 sk->sk_write_space(sk);
2011 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2013 * @skb: pointer to message buffer.
2014 * Returns true if message should be added to receive queue, false otherwise
2016 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
2018 struct sock *sk = &tsk->sk;
2019 struct net *net = sock_net(sk);
2020 struct tipc_msg *hdr = buf_msg(skb);
2021 bool con_msg = msg_connected(hdr);
2022 u32 pport = tsk_peer_port(tsk);
2023 u32 pnode = tsk_peer_node(tsk);
2024 u32 oport = msg_origport(hdr);
2025 u32 onode = msg_orignode(hdr);
2026 int err = msg_errcode(hdr);
2027 unsigned long delay;
2029 if (unlikely(msg_mcast(hdr)))
2032 switch (sk->sk_state) {
2033 case TIPC_CONNECTING:
2035 if (likely(con_msg)) {
2038 tipc_sk_finish_conn(tsk, oport, onode);
2039 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2040 /* ACK+ message with data is added to receive queue */
2041 if (msg_data_sz(hdr))
2043 /* Empty ACK-, - wake up sleeping connect() and drop */
2044 sk->sk_data_ready(sk);
2045 msg_set_dest_droppable(hdr, 1);
2048 /* Ignore connectionless message if not from listening socket */
2049 if (oport != pport || onode != pnode)
2053 if (err != TIPC_ERR_OVERLOAD)
2056 /* Prepare for new setup attempt if we have a SYN clone */
2057 if (skb_queue_empty(&sk->sk_write_queue))
2059 get_random_bytes(&delay, 2);
2060 delay %= (tsk->conn_timeout / 4);
2061 delay = msecs_to_jiffies(delay + 100);
2062 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2065 case TIPC_DISCONNECTING:
2068 /* Accept only SYN message */
2069 if (!msg_is_syn(hdr) &&
2070 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2072 if (!con_msg && !err)
2075 case TIPC_ESTABLISHED:
2076 /* Accept only connection-based messages sent by peer */
2077 if (likely(con_msg && !err && pport == oport && pnode == onode))
2079 if (!tsk_peer_msg(tsk, hdr))
2083 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2084 tipc_node_remove_conn(net, pnode, tsk->portid);
2085 sk->sk_state_change(sk);
2088 pr_err("Unknown sk_state %u\n", sk->sk_state);
2090 /* Abort connection setup attempt */
2091 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2092 sk->sk_err = ECONNREFUSED;
2093 sk->sk_state_change(sk);
2098 * rcvbuf_limit - get proper overload limit of socket receive queue
2102 * For connection oriented messages, irrespective of importance,
2103 * default queue limit is 2 MB.
2105 * For connectionless messages, queue limits are based on message
2106 * importance as follows:
2108 * TIPC_LOW_IMPORTANCE (2 MB)
2109 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2110 * TIPC_HIGH_IMPORTANCE (8 MB)
2111 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2113 * Returns overload limit according to corresponding message importance
2115 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2117 struct tipc_sock *tsk = tipc_sk(sk);
2118 struct tipc_msg *hdr = buf_msg(skb);
2120 if (unlikely(msg_in_group(hdr)))
2121 return sk->sk_rcvbuf;
2123 if (unlikely(!msg_connected(hdr)))
2124 return sk->sk_rcvbuf << msg_importance(hdr);
2126 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2127 return sk->sk_rcvbuf;
2129 return FLOWCTL_MSG_LIM;
2133 * tipc_sk_filter_rcv - validate incoming message
2135 * @skb: pointer to message.
2137 * Enqueues message on receive queue if acceptable; optionally handles
2138 * disconnect indication for a connected socket.
2140 * Called with socket lock already taken
2143 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2144 struct sk_buff_head *xmitq)
2146 bool sk_conn = !tipc_sk_type_connectionless(sk);
2147 struct tipc_sock *tsk = tipc_sk(sk);
2148 struct tipc_group *grp = tsk->group;
2149 struct tipc_msg *hdr = buf_msg(skb);
2150 struct net *net = sock_net(sk);
2151 struct sk_buff_head inputq;
2152 int limit, err = TIPC_OK;
2154 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2155 TIPC_SKB_CB(skb)->bytes_read = 0;
2156 __skb_queue_head_init(&inputq);
2157 __skb_queue_tail(&inputq, skb);
2159 if (unlikely(!msg_isdata(hdr)))
2160 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2163 tipc_group_filter_msg(grp, &inputq, xmitq);
2165 /* Validate and add to receive buffer if there is space */
2166 while ((skb = __skb_dequeue(&inputq))) {
2168 limit = rcvbuf_limit(sk, skb);
2169 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2170 (!sk_conn && msg_connected(hdr)) ||
2171 (!grp && msg_in_group(hdr)))
2172 err = TIPC_ERR_NO_PORT;
2173 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2174 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2176 atomic_inc(&sk->sk_drops);
2177 err = TIPC_ERR_OVERLOAD;
2180 if (unlikely(err)) {
2181 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2182 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2184 __skb_queue_tail(xmitq, skb);
2189 __skb_queue_tail(&sk->sk_receive_queue, skb);
2190 skb_set_owner_r(skb, sk);
2191 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2192 "rcvq >90% allocated!");
2193 sk->sk_data_ready(sk);
2198 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2202 * Caller must hold socket lock
2204 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2206 unsigned int before = sk_rmem_alloc_get(sk);
2207 struct sk_buff_head xmitq;
2210 __skb_queue_head_init(&xmitq);
2212 tipc_sk_filter_rcv(sk, skb, &xmitq);
2213 added = sk_rmem_alloc_get(sk) - before;
2214 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2216 /* Send pending response/rejected messages, if any */
2217 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2222 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2223 * inputq and try adding them to socket or backlog queue
2224 * @inputq: list of incoming buffers with potentially different destinations
2225 * @sk: socket where the buffers should be enqueued
2226 * @dport: port number for the socket
2228 * Caller must hold socket lock
2230 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2231 u32 dport, struct sk_buff_head *xmitq)
2233 unsigned long time_limit = jiffies + 2;
2234 struct sk_buff *skb;
2239 while (skb_queue_len(inputq)) {
2240 if (unlikely(time_after_eq(jiffies, time_limit)))
2243 skb = tipc_skb_dequeue(inputq, dport);
2247 /* Add message directly to receive queue if possible */
2248 if (!sock_owned_by_user(sk)) {
2249 tipc_sk_filter_rcv(sk, skb, xmitq);
2253 /* Try backlog, compensating for double-counted bytes */
2254 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2255 if (!sk->sk_backlog.len)
2256 atomic_set(dcnt, 0);
2257 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2258 if (likely(!sk_add_backlog(sk, skb, lim))) {
2259 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2260 "bklg & rcvq >90% allocated!");
2264 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2265 /* Overload => reject message back to sender */
2266 onode = tipc_own_addr(sock_net(sk));
2267 atomic_inc(&sk->sk_drops);
2268 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2269 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2271 __skb_queue_tail(xmitq, skb);
2278 * tipc_sk_rcv - handle a chain of incoming buffers
2279 * @inputq: buffer list containing the buffers
2280 * Consumes all buffers in list until inputq is empty
2281 * Note: may be called in multiple threads referring to the same queue
2283 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2285 struct sk_buff_head xmitq;
2286 u32 dnode, dport = 0;
2288 struct tipc_sock *tsk;
2290 struct sk_buff *skb;
2292 __skb_queue_head_init(&xmitq);
2293 while (skb_queue_len(inputq)) {
2294 dport = tipc_skb_peek_port(inputq, dport);
2295 tsk = tipc_sk_lookup(net, dport);
2299 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2300 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2301 spin_unlock_bh(&sk->sk_lock.slock);
2303 /* Send pending response/rejected messages, if any */
2304 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2308 /* No destination socket => dequeue skb if still there */
2309 skb = tipc_skb_dequeue(inputq, dport);
2313 /* Try secondary lookup if unresolved named message */
2314 err = TIPC_ERR_NO_PORT;
2315 if (tipc_msg_lookup_dest(net, skb, &err))
2318 /* Prepare for message rejection */
2319 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2322 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2324 dnode = msg_destnode(buf_msg(skb));
2325 tipc_node_xmit_skb(net, skb, dnode, dport);
2329 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2331 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2332 struct sock *sk = sock->sk;
2336 int err = sock_error(sk);
2341 if (signal_pending(current))
2342 return sock_intr_errno(*timeo_p);
2344 add_wait_queue(sk_sleep(sk), &wait);
2345 done = sk_wait_event(sk, timeo_p,
2346 sk->sk_state != TIPC_CONNECTING, &wait);
2347 remove_wait_queue(sk_sleep(sk), &wait);
2352 static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2354 if (addr->family != AF_TIPC)
2356 if (addr->addrtype == TIPC_SERVICE_RANGE)
2357 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2358 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2359 addr->addrtype == TIPC_SOCKET_ADDR);
2363 * tipc_connect - establish a connection to another TIPC port
2364 * @sock: socket structure
2365 * @dest: socket address for destination port
2366 * @destlen: size of socket address data structure
2367 * @flags: file-related flags associated with socket
2369 * Returns 0 on success, errno otherwise
2371 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2372 int destlen, int flags)
2374 struct sock *sk = sock->sk;
2375 struct tipc_sock *tsk = tipc_sk(sk);
2376 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2377 struct msghdr m = {NULL,};
2378 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2382 if (destlen != sizeof(struct sockaddr_tipc))
2392 if (dst->family == AF_UNSPEC) {
2393 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2394 if (!tipc_sk_type_connectionless(sk))
2398 if (!tipc_sockaddr_is_sane(dst)) {
2402 /* DGRAM/RDM connect(), just save the destaddr */
2403 if (tipc_sk_type_connectionless(sk)) {
2404 memcpy(&tsk->peer, dest, destlen);
2406 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2411 previous = sk->sk_state;
2413 switch (sk->sk_state) {
2415 /* Send a 'SYN-' to destination */
2417 m.msg_namelen = destlen;
2419 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2420 * indicate send_msg() is never blocked.
2423 m.msg_flags = MSG_DONTWAIT;
2425 res = __tipc_sendmsg(sock, &m, 0);
2426 if ((res < 0) && (res != -EWOULDBLOCK))
2429 /* Just entered TIPC_CONNECTING state; the only
2430 * difference is that return value in non-blocking
2431 * case is EINPROGRESS, rather than EALREADY.
2435 case TIPC_CONNECTING:
2437 if (previous == TIPC_CONNECTING)
2441 timeout = msecs_to_jiffies(timeout);
2442 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2443 res = tipc_wait_for_connect(sock, &timeout);
2445 case TIPC_ESTABLISHED:
2458 * tipc_listen - allow socket to listen for incoming connections
2459 * @sock: socket structure
2462 * Returns 0 on success, errno otherwise
2464 static int tipc_listen(struct socket *sock, int len)
2466 struct sock *sk = sock->sk;
2470 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2476 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2478 struct sock *sk = sock->sk;
2482 /* True wake-one mechanism for incoming connections: only
2483 * one process gets woken up, not the 'whole herd'.
2484 * Since we do not 'race & poll' for established sockets
2485 * anymore, the common case will execute the loop only once.
2488 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2489 TASK_INTERRUPTIBLE);
2490 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2492 timeo = schedule_timeout(timeo);
2496 if (!skb_queue_empty(&sk->sk_receive_queue))
2501 err = sock_intr_errno(timeo);
2502 if (signal_pending(current))
2505 finish_wait(sk_sleep(sk), &wait);
2510 * tipc_accept - wait for connection request
2511 * @sock: listening socket
2512 * @newsock: new socket that is to be connected
2513 * @flags: file-related flags associated with socket
2515 * Returns 0 on success, errno otherwise
2517 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2520 struct sock *new_sk, *sk = sock->sk;
2521 struct sk_buff *buf;
2522 struct tipc_sock *new_tsock;
2523 struct tipc_msg *msg;
2529 if (sk->sk_state != TIPC_LISTEN) {
2533 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2534 res = tipc_wait_for_accept(sock, timeo);
2538 buf = skb_peek(&sk->sk_receive_queue);
2540 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2543 security_sk_clone(sock->sk, new_sock->sk);
2545 new_sk = new_sock->sk;
2546 new_tsock = tipc_sk(new_sk);
2549 /* we lock on new_sk; but lockdep sees the lock on sk */
2550 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2553 * Reject any stray messages received by new socket
2554 * before the socket lock was taken (very, very unlikely)
2556 tsk_rej_rx_queue(new_sk);
2558 /* Connect new socket to it's peer */
2559 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2561 tsk_set_importance(new_tsock, msg_importance(msg));
2562 if (msg_named(msg)) {
2563 new_tsock->conn_type = msg_nametype(msg);
2564 new_tsock->conn_instance = msg_nameinst(msg);
2568 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2569 * Respond to 'SYN+' by queuing it on new socket.
2571 if (!msg_data_sz(msg)) {
2572 struct msghdr m = {NULL,};
2574 tsk_advance_rx_queue(sk);
2575 __tipc_sendstream(new_sock, &m, 0);
2577 __skb_dequeue(&sk->sk_receive_queue);
2578 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2579 skb_set_owner_r(buf, new_sk);
2581 release_sock(new_sk);
2588 * tipc_shutdown - shutdown socket connection
2589 * @sock: socket structure
2590 * @how: direction to close (must be SHUT_RDWR)
2592 * Terminates connection (if necessary), then purges socket's receive queue.
2594 * Returns 0 on success, errno otherwise
2596 static int tipc_shutdown(struct socket *sock, int how)
2598 struct sock *sk = sock->sk;
2601 if (how != SHUT_RDWR)
2606 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2607 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2608 sk->sk_shutdown = SEND_SHUTDOWN;
2610 if (sk->sk_state == TIPC_DISCONNECTING) {
2611 /* Discard any unreceived messages */
2612 __skb_queue_purge(&sk->sk_receive_queue);
2614 /* Wake up anyone sleeping in poll */
2615 sk->sk_state_change(sk);
2625 static void tipc_sk_check_probing_state(struct sock *sk,
2626 struct sk_buff_head *list)
2628 struct tipc_sock *tsk = tipc_sk(sk);
2629 u32 pnode = tsk_peer_node(tsk);
2630 u32 pport = tsk_peer_port(tsk);
2631 u32 self = tsk_own_node(tsk);
2632 u32 oport = tsk->portid;
2633 struct sk_buff *skb;
2635 if (tsk->probe_unacked) {
2636 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2637 sk->sk_err = ECONNABORTED;
2638 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2639 sk->sk_state_change(sk);
2642 /* Prepare new probe */
2643 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2644 pnode, self, pport, oport, TIPC_OK);
2646 __skb_queue_tail(list, skb);
2647 tsk->probe_unacked = true;
2648 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2651 static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2653 struct tipc_sock *tsk = tipc_sk(sk);
2655 /* Try again later if dest link is congested */
2656 if (tsk->cong_link_cnt) {
2657 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2660 /* Prepare SYN for retransmit */
2661 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2664 static void tipc_sk_timeout(struct timer_list *t)
2666 struct sock *sk = from_timer(sk, t, sk_timer);
2667 struct tipc_sock *tsk = tipc_sk(sk);
2668 u32 pnode = tsk_peer_node(tsk);
2669 struct sk_buff_head list;
2672 skb_queue_head_init(&list);
2675 /* Try again later if socket is busy */
2676 if (sock_owned_by_user(sk)) {
2677 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2682 if (sk->sk_state == TIPC_ESTABLISHED)
2683 tipc_sk_check_probing_state(sk, &list);
2684 else if (sk->sk_state == TIPC_CONNECTING)
2685 tipc_sk_retry_connect(sk, &list);
2689 if (!skb_queue_empty(&list))
2690 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2692 /* SYN messages may cause link congestion */
2693 if (rc == -ELINKCONG) {
2694 tipc_dest_push(&tsk->cong_links, pnode, 0);
2695 tsk->cong_link_cnt = 1;
2700 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2701 struct tipc_name_seq const *seq)
2703 struct sock *sk = &tsk->sk;
2704 struct net *net = sock_net(sk);
2705 struct publication *publ;
2708 if (scope != TIPC_NODE_SCOPE)
2709 scope = TIPC_CLUSTER_SCOPE;
2711 if (tipc_sk_connected(sk))
2713 key = tsk->portid + tsk->pub_count + 1;
2714 if (key == tsk->portid)
2717 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2718 scope, tsk->portid, key);
2719 if (unlikely(!publ))
2722 list_add(&publ->binding_sock, &tsk->publications);
2728 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2729 struct tipc_name_seq const *seq)
2731 struct net *net = sock_net(&tsk->sk);
2732 struct publication *publ;
2733 struct publication *safe;
2736 if (scope != TIPC_NODE_SCOPE)
2737 scope = TIPC_CLUSTER_SCOPE;
2739 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2741 if (publ->scope != scope)
2743 if (publ->type != seq->type)
2745 if (publ->lower != seq->lower)
2747 if (publ->upper != seq->upper)
2749 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2750 publ->upper, publ->key);
2754 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2755 publ->upper, publ->key);
2758 if (list_empty(&tsk->publications))
2763 /* tipc_sk_reinit: set non-zero address in all existing sockets
2764 * when we go from standalone to network mode.
2766 void tipc_sk_reinit(struct net *net)
2768 struct tipc_net *tn = net_generic(net, tipc_net_id);
2769 struct rhashtable_iter iter;
2770 struct tipc_sock *tsk;
2771 struct tipc_msg *msg;
2773 rhashtable_walk_enter(&tn->sk_rht, &iter);
2776 rhashtable_walk_start(&iter);
2778 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2779 sock_hold(&tsk->sk);
2780 rhashtable_walk_stop(&iter);
2781 lock_sock(&tsk->sk);
2783 msg_set_prevnode(msg, tipc_own_addr(net));
2784 msg_set_orignode(msg, tipc_own_addr(net));
2785 release_sock(&tsk->sk);
2786 rhashtable_walk_start(&iter);
2790 rhashtable_walk_stop(&iter);
2791 } while (tsk == ERR_PTR(-EAGAIN));
2793 rhashtable_walk_exit(&iter);
2796 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2798 struct tipc_net *tn = net_generic(net, tipc_net_id);
2799 struct tipc_sock *tsk;
2802 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2804 sock_hold(&tsk->sk);
2810 static int tipc_sk_insert(struct tipc_sock *tsk)
2812 struct sock *sk = &tsk->sk;
2813 struct net *net = sock_net(sk);
2814 struct tipc_net *tn = net_generic(net, tipc_net_id);
2815 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2816 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2818 while (remaining--) {
2820 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2821 portid = TIPC_MIN_PORT;
2822 tsk->portid = portid;
2823 sock_hold(&tsk->sk);
2824 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2833 static void tipc_sk_remove(struct tipc_sock *tsk)
2835 struct sock *sk = &tsk->sk;
2836 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2838 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2839 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2844 static const struct rhashtable_params tsk_rht_params = {
2846 .head_offset = offsetof(struct tipc_sock, node),
2847 .key_offset = offsetof(struct tipc_sock, portid),
2848 .key_len = sizeof(u32), /* portid */
2849 .max_size = 1048576,
2851 .automatic_shrinking = true,
2854 int tipc_sk_rht_init(struct net *net)
2856 struct tipc_net *tn = net_generic(net, tipc_net_id);
2858 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2861 void tipc_sk_rht_destroy(struct net *net)
2863 struct tipc_net *tn = net_generic(net, tipc_net_id);
2865 /* Wait for socket readers to complete */
2868 rhashtable_destroy(&tn->sk_rht);
2871 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2873 struct net *net = sock_net(&tsk->sk);
2874 struct tipc_group *grp = tsk->group;
2875 struct tipc_msg *hdr = &tsk->phdr;
2876 struct tipc_name_seq seq;
2879 if (mreq->type < TIPC_RESERVED_TYPES)
2881 if (mreq->scope > TIPC_NODE_SCOPE)
2885 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2889 msg_set_lookup_scope(hdr, mreq->scope);
2890 msg_set_nametype(hdr, mreq->type);
2891 msg_set_dest_droppable(hdr, true);
2892 seq.type = mreq->type;
2893 seq.lower = mreq->instance;
2894 seq.upper = seq.lower;
2895 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2896 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2898 tipc_group_delete(net, grp);
2902 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2903 tsk->mc_method.rcast = true;
2904 tsk->mc_method.mandatory = true;
2905 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2909 static int tipc_sk_leave(struct tipc_sock *tsk)
2911 struct net *net = sock_net(&tsk->sk);
2912 struct tipc_group *grp = tsk->group;
2913 struct tipc_name_seq seq;
2918 tipc_group_self(grp, &seq, &scope);
2919 tipc_group_delete(net, grp);
2921 tipc_sk_withdraw(tsk, scope, &seq);
2926 * tipc_setsockopt - set socket option
2927 * @sock: socket structure
2928 * @lvl: option level
2929 * @opt: option identifier
2930 * @ov: pointer to new option value
2931 * @ol: length of option value
2933 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2934 * (to ease compatibility).
2936 * Returns 0 on success, errno otherwise
2938 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2939 char __user *ov, unsigned int ol)
2941 struct sock *sk = sock->sk;
2942 struct tipc_sock *tsk = tipc_sk(sk);
2943 struct tipc_group_req mreq;
2947 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2949 if (lvl != SOL_TIPC)
2950 return -ENOPROTOOPT;
2953 case TIPC_IMPORTANCE:
2954 case TIPC_SRC_DROPPABLE:
2955 case TIPC_DEST_DROPPABLE:
2956 case TIPC_CONN_TIMEOUT:
2957 if (ol < sizeof(value))
2959 if (get_user(value, (u32 __user *)ov))
2962 case TIPC_GROUP_JOIN:
2963 if (ol < sizeof(mreq))
2965 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2976 case TIPC_IMPORTANCE:
2977 res = tsk_set_importance(tsk, value);
2979 case TIPC_SRC_DROPPABLE:
2980 if (sock->type != SOCK_STREAM)
2981 tsk_set_unreliable(tsk, value);
2985 case TIPC_DEST_DROPPABLE:
2986 tsk_set_unreturnable(tsk, value);
2988 case TIPC_CONN_TIMEOUT:
2989 tipc_sk(sk)->conn_timeout = value;
2991 case TIPC_MCAST_BROADCAST:
2992 tsk->mc_method.rcast = false;
2993 tsk->mc_method.mandatory = true;
2995 case TIPC_MCAST_REPLICAST:
2996 tsk->mc_method.rcast = true;
2997 tsk->mc_method.mandatory = true;
2999 case TIPC_GROUP_JOIN:
3000 res = tipc_sk_join(tsk, &mreq);
3002 case TIPC_GROUP_LEAVE:
3003 res = tipc_sk_leave(tsk);
3015 * tipc_getsockopt - get socket option
3016 * @sock: socket structure
3017 * @lvl: option level
3018 * @opt: option identifier
3019 * @ov: receptacle for option value
3020 * @ol: receptacle for length of option value
3022 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3023 * (to ease compatibility).
3025 * Returns 0 on success, errno otherwise
3027 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3028 char __user *ov, int __user *ol)
3030 struct sock *sk = sock->sk;
3031 struct tipc_sock *tsk = tipc_sk(sk);
3032 struct tipc_name_seq seq;
3037 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3038 return put_user(0, ol);
3039 if (lvl != SOL_TIPC)
3040 return -ENOPROTOOPT;
3041 res = get_user(len, ol);
3048 case TIPC_IMPORTANCE:
3049 value = tsk_importance(tsk);
3051 case TIPC_SRC_DROPPABLE:
3052 value = tsk_unreliable(tsk);
3054 case TIPC_DEST_DROPPABLE:
3055 value = tsk_unreturnable(tsk);
3057 case TIPC_CONN_TIMEOUT:
3058 value = tsk->conn_timeout;
3059 /* no need to set "res", since already 0 at this point */
3061 case TIPC_NODE_RECVQ_DEPTH:
3062 value = 0; /* was tipc_queue_size, now obsolete */
3064 case TIPC_SOCK_RECVQ_DEPTH:
3065 value = skb_queue_len(&sk->sk_receive_queue);
3067 case TIPC_GROUP_JOIN:
3070 tipc_group_self(tsk->group, &seq, &scope);
3080 return res; /* "get" failed */
3082 if (len < sizeof(value))
3085 if (copy_to_user(ov, &value, sizeof(value)))
3088 return put_user(sizeof(value), ol);
3091 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3093 struct net *net = sock_net(sock->sk);
3094 struct tipc_sioc_nodeid_req nr = {0};
3095 struct tipc_sioc_ln_req lnr;
3096 void __user *argp = (void __user *)arg;
3099 case SIOCGETLINKNAME:
3100 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3102 if (!tipc_node_get_linkname(net,
3103 lnr.bearer_id & 0xffff, lnr.peer,
3104 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3105 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3109 return -EADDRNOTAVAIL;
3111 if (copy_from_user(&nr, argp, sizeof(nr)))
3113 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3114 return -EADDRNOTAVAIL;
3115 if (copy_to_user(argp, &nr, sizeof(nr)))
3119 return -ENOIOCTLCMD;
3123 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3125 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3126 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3127 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3129 tsk1->peer.family = AF_TIPC;
3130 tsk1->peer.addrtype = TIPC_ADDR_ID;
3131 tsk1->peer.scope = TIPC_NODE_SCOPE;
3132 tsk1->peer.addr.id.ref = tsk2->portid;
3133 tsk1->peer.addr.id.node = onode;
3134 tsk2->peer.family = AF_TIPC;
3135 tsk2->peer.addrtype = TIPC_ADDR_ID;
3136 tsk2->peer.scope = TIPC_NODE_SCOPE;
3137 tsk2->peer.addr.id.ref = tsk1->portid;
3138 tsk2->peer.addr.id.node = onode;
3140 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3141 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3145 /* Protocol switches for the various types of TIPC sockets */
3147 static const struct proto_ops msg_ops = {
3148 .owner = THIS_MODULE,
3150 .release = tipc_release,
3152 .connect = tipc_connect,
3153 .socketpair = tipc_socketpair,
3154 .accept = sock_no_accept,
3155 .getname = tipc_getname,
3157 .ioctl = tipc_ioctl,
3158 .listen = sock_no_listen,
3159 .shutdown = tipc_shutdown,
3160 .setsockopt = tipc_setsockopt,
3161 .getsockopt = tipc_getsockopt,
3162 .sendmsg = tipc_sendmsg,
3163 .recvmsg = tipc_recvmsg,
3164 .mmap = sock_no_mmap,
3165 .sendpage = sock_no_sendpage
3168 static const struct proto_ops packet_ops = {
3169 .owner = THIS_MODULE,
3171 .release = tipc_release,
3173 .connect = tipc_connect,
3174 .socketpair = tipc_socketpair,
3175 .accept = tipc_accept,
3176 .getname = tipc_getname,
3178 .ioctl = tipc_ioctl,
3179 .listen = tipc_listen,
3180 .shutdown = tipc_shutdown,
3181 .setsockopt = tipc_setsockopt,
3182 .getsockopt = tipc_getsockopt,
3183 .sendmsg = tipc_send_packet,
3184 .recvmsg = tipc_recvmsg,
3185 .mmap = sock_no_mmap,
3186 .sendpage = sock_no_sendpage
3189 static const struct proto_ops stream_ops = {
3190 .owner = THIS_MODULE,
3192 .release = tipc_release,
3194 .connect = tipc_connect,
3195 .socketpair = tipc_socketpair,
3196 .accept = tipc_accept,
3197 .getname = tipc_getname,
3199 .ioctl = tipc_ioctl,
3200 .listen = tipc_listen,
3201 .shutdown = tipc_shutdown,
3202 .setsockopt = tipc_setsockopt,
3203 .getsockopt = tipc_getsockopt,
3204 .sendmsg = tipc_sendstream,
3205 .recvmsg = tipc_recvstream,
3206 .mmap = sock_no_mmap,
3207 .sendpage = sock_no_sendpage
3210 static const struct net_proto_family tipc_family_ops = {
3211 .owner = THIS_MODULE,
3213 .create = tipc_sk_create
3216 static struct proto tipc_proto = {
3218 .owner = THIS_MODULE,
3219 .obj_size = sizeof(struct tipc_sock),
3220 .sysctl_rmem = sysctl_tipc_rmem
3224 * tipc_socket_init - initialize TIPC socket interface
3226 * Returns 0 on success, errno otherwise
3228 int tipc_socket_init(void)
3232 res = proto_register(&tipc_proto, 1);
3234 pr_err("Failed to register TIPC protocol type\n");
3238 res = sock_register(&tipc_family_ops);
3240 pr_err("Failed to register TIPC socket type\n");
3241 proto_unregister(&tipc_proto);
3249 * tipc_socket_stop - stop TIPC socket interface
3251 void tipc_socket_stop(void)
3253 sock_unregister(tipc_family_ops.family);
3254 proto_unregister(&tipc_proto);
3257 /* Caller should hold socket lock for the passed tipc socket. */
3258 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3262 struct nlattr *nest;
3264 peer_node = tsk_peer_node(tsk);
3265 peer_port = tsk_peer_port(tsk);
3267 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3271 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3273 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3276 if (tsk->conn_type != 0) {
3277 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3279 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3281 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3284 nla_nest_end(skb, nest);
3289 nla_nest_cancel(skb, nest);
3294 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3297 struct net *net = sock_net(skb->sk);
3298 struct sock *sk = &tsk->sk;
3300 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3301 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3304 if (tipc_sk_connected(sk)) {
3305 if (__tipc_nl_add_sk_con(skb, tsk))
3307 } else if (!list_empty(&tsk->publications)) {
3308 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3314 /* Caller should hold socket lock for the passed tipc socket. */
3315 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3316 struct tipc_sock *tsk)
3318 struct nlattr *attrs;
3321 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3322 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3326 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3328 goto genlmsg_cancel;
3330 if (__tipc_nl_add_sk_info(skb, tsk))
3331 goto attr_msg_cancel;
3333 nla_nest_end(skb, attrs);
3334 genlmsg_end(skb, hdr);
3339 nla_nest_cancel(skb, attrs);
3341 genlmsg_cancel(skb, hdr);
3346 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3347 int (*skb_handler)(struct sk_buff *skb,
3348 struct netlink_callback *cb,
3349 struct tipc_sock *tsk))
3351 struct rhashtable_iter *iter = (void *)cb->args[4];
3352 struct tipc_sock *tsk;
3355 rhashtable_walk_start(iter);
3356 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3359 if (err == -EAGAIN) {
3366 sock_hold(&tsk->sk);
3367 rhashtable_walk_stop(iter);
3368 lock_sock(&tsk->sk);
3369 err = skb_handler(skb, cb, tsk);
3371 release_sock(&tsk->sk);
3375 release_sock(&tsk->sk);
3376 rhashtable_walk_start(iter);
3379 rhashtable_walk_stop(iter);
3383 EXPORT_SYMBOL(tipc_nl_sk_walk);
3385 int tipc_dump_start(struct netlink_callback *cb)
3387 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3389 EXPORT_SYMBOL(tipc_dump_start);
3391 int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3393 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3394 struct rhashtable_iter *iter = (void *)cb->args[4];
3395 struct tipc_net *tn = tipc_net(net);
3398 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3402 cb->args[4] = (long)iter;
3405 rhashtable_walk_enter(&tn->sk_rht, iter);
3409 int tipc_dump_done(struct netlink_callback *cb)
3411 struct rhashtable_iter *hti = (void *)cb->args[4];
3413 rhashtable_walk_exit(hti);
3417 EXPORT_SYMBOL(tipc_dump_done);
3419 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3420 struct tipc_sock *tsk, u32 sk_filter_state,
3421 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3423 struct sock *sk = &tsk->sk;
3424 struct nlattr *attrs;
3425 struct nlattr *stat;
3427 /*filter response w.r.t sk_state*/
3428 if (!(sk_filter_state & (1 << sk->sk_state)))
3431 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3435 if (__tipc_nl_add_sk_info(skb, tsk))
3436 goto attr_msg_cancel;
3438 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3439 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3440 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3441 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3442 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3444 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3445 tipc_diag_gen_cookie(sk),
3447 goto attr_msg_cancel;
3449 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3451 goto attr_msg_cancel;
3453 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3454 skb_queue_len(&sk->sk_receive_queue)) ||
3455 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3456 skb_queue_len(&sk->sk_write_queue)) ||
3457 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3458 atomic_read(&sk->sk_drops)))
3459 goto stat_msg_cancel;
3461 if (tsk->cong_link_cnt &&
3462 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3463 goto stat_msg_cancel;
3465 if (tsk_conn_cong(tsk) &&
3466 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3467 goto stat_msg_cancel;
3469 nla_nest_end(skb, stat);
3472 if (tipc_group_fill_sock_diag(tsk->group, skb))
3473 goto stat_msg_cancel;
3475 nla_nest_end(skb, attrs);
3480 nla_nest_cancel(skb, stat);
3482 nla_nest_cancel(skb, attrs);
3486 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3488 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3490 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3493 /* Caller should hold socket lock for the passed tipc socket. */
3494 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3495 struct netlink_callback *cb,
3496 struct publication *publ)
3499 struct nlattr *attrs;
3501 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3502 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3506 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3508 goto genlmsg_cancel;
3510 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3511 goto attr_msg_cancel;
3512 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3513 goto attr_msg_cancel;
3514 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3515 goto attr_msg_cancel;
3516 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3517 goto attr_msg_cancel;
3519 nla_nest_end(skb, attrs);
3520 genlmsg_end(skb, hdr);
3525 nla_nest_cancel(skb, attrs);
3527 genlmsg_cancel(skb, hdr);
3532 /* Caller should hold socket lock for the passed tipc socket. */
3533 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3534 struct netlink_callback *cb,
3535 struct tipc_sock *tsk, u32 *last_publ)
3538 struct publication *p;
3541 list_for_each_entry(p, &tsk->publications, binding_sock) {
3542 if (p->key == *last_publ)
3545 if (p->key != *last_publ) {
3546 /* We never set seq or call nl_dump_check_consistent()
3547 * this means that setting prev_seq here will cause the
3548 * consistence check to fail in the netlink callback
3549 * handler. Resulting in the last NLMSG_DONE message
3550 * having the NLM_F_DUMP_INTR flag set.
3557 p = list_first_entry(&tsk->publications, struct publication,
3561 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3562 err = __tipc_nl_add_sk_publ(skb, cb, p);
3564 *last_publ = p->key;
3573 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3576 u32 tsk_portid = cb->args[0];
3577 u32 last_publ = cb->args[1];
3578 u32 done = cb->args[2];
3579 struct net *net = sock_net(skb->sk);
3580 struct tipc_sock *tsk;
3583 struct nlattr **attrs;
3584 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3586 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3590 if (!attrs[TIPC_NLA_SOCK])
3593 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3594 attrs[TIPC_NLA_SOCK],
3595 tipc_nl_sock_policy, NULL);
3599 if (!sock[TIPC_NLA_SOCK_REF])
3602 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3608 tsk = tipc_sk_lookup(net, tsk_portid);
3612 lock_sock(&tsk->sk);
3613 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3616 release_sock(&tsk->sk);
3619 cb->args[0] = tsk_portid;
3620 cb->args[1] = last_publ;
3627 * tipc_sk_filtering - check if a socket should be traced
3628 * @sk: the socket to be examined
3629 * @sysctl_tipc_sk_filter[]: the socket tuple for filtering,
3630 * (portid, sock type, name type, name lower, name upper)
3632 * Returns true if the socket meets the socket tuple data
3633 * (value 0 = 'any') or when there is no tuple set (all = 0),
3636 bool tipc_sk_filtering(struct sock *sk)
3638 struct tipc_sock *tsk;
3639 struct publication *p;
3640 u32 _port, _sktype, _type, _lower, _upper;
3641 u32 type = 0, lower = 0, upper = 0;
3648 _port = sysctl_tipc_sk_filter[0];
3649 _sktype = sysctl_tipc_sk_filter[1];
3650 _type = sysctl_tipc_sk_filter[2];
3651 _lower = sysctl_tipc_sk_filter[3];
3652 _upper = sysctl_tipc_sk_filter[4];
3654 if (!_port && !_sktype && !_type && !_lower && !_upper)
3658 return (_port == tsk->portid);
3660 if (_sktype && _sktype != sk->sk_type)
3663 if (tsk->published) {
3664 p = list_first_entry_or_null(&tsk->publications,
3665 struct publication, binding_sock);
3673 if (!tipc_sk_type_connectionless(sk)) {
3674 type = tsk->conn_type;
3675 lower = tsk->conn_instance;
3676 upper = tsk->conn_instance;
3679 if ((_type && _type != type) || (_lower && _lower != lower) ||
3680 (_upper && _upper != upper))
3686 u32 tipc_sock_get_portid(struct sock *sk)
3688 return (sk) ? (tipc_sk(sk))->portid : 0;
3692 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3693 * both the rcv and backlog queues are considered
3694 * @sk: tipc sk to be checked
3695 * @skb: tipc msg to be checked
3697 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3700 bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3702 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3703 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3704 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3706 return (qsize > lim * 90 / 100);
3710 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3711 * only the rcv queue is considered
3712 * @sk: tipc sk to be checked
3713 * @skb: tipc msg to be checked
3715 * Returns true if the socket rx queue allocation is > 90%, otherwise false
3718 bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3720 unsigned int lim = rcvbuf_limit(sk, skb);
3721 unsigned int qsize = sk_rmem_alloc_get(sk);
3723 return (qsize > lim * 90 / 100);
3727 * tipc_sk_dump - dump TIPC socket
3728 * @sk: tipc sk to be dumped
3729 * @dqueues: bitmask to decide if any socket queue to be dumped?
3730 * - TIPC_DUMP_NONE: don't dump socket queues
3731 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3732 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3733 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3734 * - TIPC_DUMP_ALL: dump all the socket queues above
3735 * @buf: returned buffer of dump data in format
3737 int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3740 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3741 struct tipc_sock *tsk;
3742 struct publication *p;
3746 i += scnprintf(buf, sz, "sk data: (null)\n");
3751 tsk_connected = !tipc_sk_type_connectionless(sk);
3753 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3754 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3755 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3756 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3757 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3758 if (tsk_connected) {
3759 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3760 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3761 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_type);
3762 i += scnprintf(buf + i, sz - i, " %u", tsk->conn_instance);
3764 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3765 if (tsk->published) {
3766 p = list_first_entry_or_null(&tsk->publications,
3767 struct publication, binding_sock);
3768 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->type : 0);
3769 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->lower : 0);
3770 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->upper : 0);
3772 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3773 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3774 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3775 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3776 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3777 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3778 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3779 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3780 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3781 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3782 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3783 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3784 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3785 i += scnprintf(buf + i, sz - i, " | %d\n", sk->sk_backlog.len);
3787 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3788 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3789 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3792 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3793 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3794 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3797 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
3798 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
3799 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
3800 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
3801 i += scnprintf(buf + i, sz - i, " tail ");
3802 i += tipc_skb_dump(sk->sk_backlog.tail, false,