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"
50 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
51 #define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
52 #define TIPC_FWD_MSG 1
53 #define TIPC_MAX_PORT 0xffffffff
54 #define TIPC_MIN_PORT 1
55 #define TIPC_ACK_RATE 4 /* ACK at 1/4 of of rcv window size */
58 TIPC_LISTEN = TCP_LISTEN,
59 TIPC_ESTABLISHED = TCP_ESTABLISHED,
60 TIPC_OPEN = TCP_CLOSE,
61 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
62 TIPC_CONNECTING = TCP_SYN_SENT,
65 struct sockaddr_pair {
66 struct sockaddr_tipc sock;
67 struct sockaddr_tipc member;
71 * struct tipc_sock - TIPC socket structure
72 * @sk: socket - interacts with 'port' and with user via the socket API
73 * @conn_type: TIPC type used when connection was established
74 * @conn_instance: TIPC instance used when connection was established
75 * @published: non-zero if port has one or more associated names
76 * @max_pkt: maximum packet size "hint" used when building messages sent by port
77 * @portid: unique port identity in TIPC socket hash table
78 * @phdr: preformatted message header used when sending messages
79 * #cong_links: list of congested links
80 * @publications: list of publications for port
81 * @blocking_link: address of the congested link we are currently sleeping on
82 * @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;
106 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 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
240 /* tipc_sk_respond() : send response message back to sender
242 static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
246 u32 onode = tipc_own_addr(sock_net(sk));
248 if (!tipc_msg_reverse(onode, &skb, err))
251 dnode = msg_destnode(buf_msg(skb));
252 selector = msg_origport(buf_msg(skb));
253 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
257 * tsk_rej_rx_queue - reject all buffers in socket receive queue
259 * Caller must hold socket lock
261 static void tsk_rej_rx_queue(struct sock *sk)
265 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
266 tipc_sk_respond(sk, skb, TIPC_ERR_NO_PORT);
269 static bool tipc_sk_connected(struct sock *sk)
271 return sk->sk_state == TIPC_ESTABLISHED;
274 /* tipc_sk_type_connectionless - check if the socket is datagram socket
277 * Returns true if connection less, false otherwise
279 static bool tipc_sk_type_connectionless(struct sock *sk)
281 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
284 /* tsk_peer_msg - verify if message was sent by connected port's peer
286 * Handles cases where the node's network address has changed from
287 * the default of <0.0.0> to its configured setting.
289 static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
291 struct sock *sk = &tsk->sk;
292 u32 self = tipc_own_addr(sock_net(sk));
293 u32 peer_port = tsk_peer_port(tsk);
294 u32 orig_node, peer_node;
296 if (unlikely(!tipc_sk_connected(sk)))
299 if (unlikely(msg_origport(msg) != peer_port))
302 orig_node = msg_orignode(msg);
303 peer_node = tsk_peer_node(tsk);
305 if (likely(orig_node == peer_node))
308 if (!orig_node && peer_node == self)
311 if (!peer_node && orig_node == self)
317 /* tipc_set_sk_state - set the sk_state of the socket
320 * Caller must hold socket lock
322 * Returns 0 on success, errno otherwise
324 static int tipc_set_sk_state(struct sock *sk, int state)
326 int oldsk_state = sk->sk_state;
334 case TIPC_CONNECTING:
335 if (oldsk_state == TIPC_OPEN)
338 case TIPC_ESTABLISHED:
339 if (oldsk_state == TIPC_CONNECTING ||
340 oldsk_state == TIPC_OPEN)
343 case TIPC_DISCONNECTING:
344 if (oldsk_state == TIPC_CONNECTING ||
345 oldsk_state == TIPC_ESTABLISHED)
351 sk->sk_state = state;
356 static int tipc_sk_sock_err(struct socket *sock, long *timeout)
358 struct sock *sk = sock->sk;
359 int err = sock_error(sk);
360 int typ = sock->type;
364 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
365 if (sk->sk_state == TIPC_DISCONNECTING)
367 else if (!tipc_sk_connected(sk))
372 if (signal_pending(current))
373 return sock_intr_errno(*timeout);
378 #define tipc_wait_for_cond(sock_, timeo_, condition_) \
383 while ((rc_ = !(condition_))) { \
384 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
386 rc_ = tipc_sk_sock_err((sock_), timeo_); \
389 prepare_to_wait(sk_sleep(sk_), &wait_, TASK_INTERRUPTIBLE); \
391 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
392 sched_annotate_sleep(); \
394 remove_wait_queue(sk_sleep(sk_), &wait_); \
400 * tipc_sk_create - create a TIPC socket
401 * @net: network namespace (must be default network)
402 * @sock: pre-allocated socket structure
403 * @protocol: protocol indicator (must be 0)
404 * @kern: caused by kernel or by userspace?
406 * This routine creates additional data structures used by the TIPC socket,
407 * initializes them, and links them together.
409 * Returns 0 on success, errno otherwise
411 static int tipc_sk_create(struct net *net, struct socket *sock,
412 int protocol, int kern)
415 const struct proto_ops *ops;
417 struct tipc_sock *tsk;
418 struct tipc_msg *msg;
420 /* Validate arguments */
421 if (unlikely(protocol != 0))
422 return -EPROTONOSUPPORT;
424 switch (sock->type) {
439 /* Allocate socket's protocol area */
440 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
445 tsk->max_pkt = MAX_PKT_DEFAULT;
446 INIT_LIST_HEAD(&tsk->publications);
447 INIT_LIST_HEAD(&tsk->cong_links);
449 tn = net_generic(sock_net(sk), tipc_net_id);
451 /* Finish initializing socket data structures */
453 sock_init_data(sock, sk);
454 tipc_set_sk_state(sk, TIPC_OPEN);
455 if (tipc_sk_insert(tsk)) {
456 pr_warn("Socket create failed; port number exhausted\n");
460 /* Ensure tsk is visible before we read own_addr. */
463 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
464 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
466 msg_set_origport(msg, tsk->portid);
467 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
469 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
470 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
471 sk->sk_data_ready = tipc_data_ready;
472 sk->sk_write_space = tipc_write_space;
473 sk->sk_destruct = tipc_sock_destruct;
474 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
475 tsk->group_is_open = true;
476 atomic_set(&tsk->dupl_rcvcnt, 0);
478 /* Start out with safe limits until we receive an advertised window */
479 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
480 tsk->rcv_win = tsk->snd_win;
482 if (tipc_sk_type_connectionless(sk)) {
483 tsk_set_unreturnable(tsk, true);
484 if (sock->type == SOCK_DGRAM)
485 tsk_set_unreliable(tsk, true);
491 static void tipc_sk_callback(struct rcu_head *head)
493 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
498 /* Caller should hold socket lock for the socket. */
499 static void __tipc_shutdown(struct socket *sock, int error)
501 struct sock *sk = sock->sk;
502 struct tipc_sock *tsk = tipc_sk(sk);
503 struct net *net = sock_net(sk);
504 long timeout = CONN_TIMEOUT_DEFAULT;
505 u32 dnode = tsk_peer_node(tsk);
508 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
509 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
510 !tsk_conn_cong(tsk)));
512 /* Reject all unreceived messages, except on an active connection
513 * (which disconnects locally & sends a 'FIN+' to peer).
515 while ((skb = __skb_dequeue(&sk->sk_receive_queue)) != NULL) {
516 if (TIPC_SKB_CB(skb)->bytes_read) {
520 if (!tipc_sk_type_connectionless(sk) &&
521 sk->sk_state != TIPC_DISCONNECTING) {
522 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
523 tipc_node_remove_conn(net, dnode, tsk->portid);
525 tipc_sk_respond(sk, skb, error);
528 if (tipc_sk_type_connectionless(sk))
531 if (sk->sk_state != TIPC_DISCONNECTING) {
532 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
533 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
534 tsk_own_node(tsk), tsk_peer_port(tsk),
537 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
538 tipc_node_remove_conn(net, dnode, tsk->portid);
539 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
544 * tipc_release - destroy a TIPC socket
545 * @sock: socket to destroy
547 * This routine cleans up any messages that are still queued on the socket.
548 * For DGRAM and RDM socket types, all queued messages are rejected.
549 * For SEQPACKET and STREAM socket types, the first message is rejected
550 * and any others are discarded. (If the first message on a STREAM socket
551 * is partially-read, it is discarded and the next one is rejected instead.)
553 * NOTE: Rejected messages are not necessarily returned to the sender! They
554 * are returned or discarded according to the "destination droppable" setting
555 * specified for the message by the sender.
557 * Returns 0 on success, errno otherwise
559 static int tipc_release(struct socket *sock)
561 struct sock *sk = sock->sk;
562 struct tipc_sock *tsk;
565 * Exit if socket isn't fully initialized (occurs when a failed accept()
566 * releases a pre-allocated child socket that was never used)
574 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
575 sk->sk_shutdown = SHUTDOWN_MASK;
577 tipc_sk_withdraw(tsk, 0, NULL);
578 sk_stop_timer(sk, &sk->sk_timer);
581 /* Reject any messages that accumulated in backlog queue */
583 tipc_dest_list_purge(&tsk->cong_links);
584 tsk->cong_link_cnt = 0;
585 call_rcu(&tsk->rcu, tipc_sk_callback);
592 * tipc_bind - associate or disassocate TIPC name(s) with a socket
593 * @sock: socket structure
594 * @uaddr: socket address describing name(s) and desired operation
595 * @uaddr_len: size of socket address data structure
597 * Name and name sequence binding is indicated using a positive scope value;
598 * a negative scope value unbinds the specified name. Specifying no name
599 * (i.e. a socket address length of 0) unbinds all names from the socket.
601 * Returns 0 on success, errno otherwise
603 * NOTE: This routine doesn't need to take the socket lock since it doesn't
604 * access any non-constant socket information.
606 static int tipc_bind(struct socket *sock, struct sockaddr *uaddr,
609 struct sock *sk = sock->sk;
610 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
611 struct tipc_sock *tsk = tipc_sk(sk);
615 if (unlikely(!uaddr_len)) {
616 res = tipc_sk_withdraw(tsk, 0, NULL);
623 if (uaddr_len < sizeof(struct sockaddr_tipc)) {
627 if (addr->family != AF_TIPC) {
632 if (addr->addrtype == TIPC_ADDR_NAME)
633 addr->addr.nameseq.upper = addr->addr.nameseq.lower;
634 else if (addr->addrtype != TIPC_ADDR_NAMESEQ) {
639 if ((addr->addr.nameseq.type < TIPC_RESERVED_TYPES) &&
640 (addr->addr.nameseq.type != TIPC_TOP_SRV) &&
641 (addr->addr.nameseq.type != TIPC_CFG_SRV)) {
646 res = (addr->scope >= 0) ?
647 tipc_sk_publish(tsk, addr->scope, &addr->addr.nameseq) :
648 tipc_sk_withdraw(tsk, -addr->scope, &addr->addr.nameseq);
655 * tipc_getname - get port ID of socket or peer socket
656 * @sock: socket structure
657 * @uaddr: area for returned socket address
658 * @uaddr_len: area for returned length of socket address
659 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
661 * Returns 0 on success, errno otherwise
663 * NOTE: This routine doesn't need to take the socket lock since it only
664 * accesses socket information that is unchanging (or which changes in
665 * a completely predictable manner).
667 static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
670 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
671 struct sock *sk = sock->sk;
672 struct tipc_sock *tsk = tipc_sk(sk);
674 memset(addr, 0, sizeof(*addr));
676 if ((!tipc_sk_connected(sk)) &&
677 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
679 addr->addr.id.ref = tsk_peer_port(tsk);
680 addr->addr.id.node = tsk_peer_node(tsk);
682 addr->addr.id.ref = tsk->portid;
683 addr->addr.id.node = tipc_own_addr(sock_net(sk));
686 addr->addrtype = TIPC_ADDR_ID;
687 addr->family = AF_TIPC;
689 addr->addr.name.domain = 0;
691 return sizeof(*addr);
695 * tipc_poll - read and possibly block on pollmask
696 * @file: file structure associated with the socket
697 * @sock: socket for which to calculate the poll bits
700 * Returns pollmask value
703 * It appears that the usual socket locking mechanisms are not useful here
704 * since the pollmask info is potentially out-of-date the moment this routine
705 * exits. TCP and other protocols seem to rely on higher level poll routines
706 * to handle any preventable race conditions, so TIPC will do the same ...
708 * IMPORTANT: The fact that a read or write operation is indicated does NOT
709 * imply that the operation will succeed, merely that it should be performed
710 * and will not block.
712 static __poll_t tipc_poll(struct file *file, struct socket *sock,
715 struct sock *sk = sock->sk;
716 struct tipc_sock *tsk = tipc_sk(sk);
717 __poll_t revents = 0;
719 sock_poll_wait(file, wait);
721 if (sk->sk_shutdown & RCV_SHUTDOWN)
722 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
723 if (sk->sk_shutdown == SHUTDOWN_MASK)
726 switch (sk->sk_state) {
727 case TIPC_ESTABLISHED:
728 case TIPC_CONNECTING:
729 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
733 if (!skb_queue_empty(&sk->sk_receive_queue))
734 revents |= EPOLLIN | EPOLLRDNORM;
737 if (tsk->group_is_open && !tsk->cong_link_cnt)
739 if (!tipc_sk_type_connectionless(sk))
741 if (skb_queue_empty(&sk->sk_receive_queue))
743 revents |= EPOLLIN | EPOLLRDNORM;
745 case TIPC_DISCONNECTING:
746 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
753 * tipc_sendmcast - send multicast message
754 * @sock: socket structure
755 * @seq: destination address
756 * @msg: message to send
757 * @dlen: length of data to send
758 * @timeout: timeout to wait for wakeup
760 * Called from function tipc_sendmsg(), which has done all sanity checks
761 * Returns the number of bytes sent on success, or errno
763 static int tipc_sendmcast(struct socket *sock, struct tipc_name_seq *seq,
764 struct msghdr *msg, size_t dlen, long timeout)
766 struct sock *sk = sock->sk;
767 struct tipc_sock *tsk = tipc_sk(sk);
768 struct tipc_msg *hdr = &tsk->phdr;
769 struct net *net = sock_net(sk);
770 int mtu = tipc_bcast_get_mtu(net);
771 struct tipc_mc_method *method = &tsk->mc_method;
772 struct sk_buff_head pkts;
773 struct tipc_nlist dsts;
779 /* Block or return if any destination link is congested */
780 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
784 /* Lookup destination nodes */
785 tipc_nlist_init(&dsts, tipc_own_addr(net));
786 tipc_nametbl_lookup_dst_nodes(net, seq->type, seq->lower,
788 if (!dsts.local && !dsts.remote)
789 return -EHOSTUNREACH;
791 /* Build message header */
792 msg_set_type(hdr, TIPC_MCAST_MSG);
793 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
794 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
795 msg_set_destport(hdr, 0);
796 msg_set_destnode(hdr, 0);
797 msg_set_nametype(hdr, seq->type);
798 msg_set_namelower(hdr, seq->lower);
799 msg_set_nameupper(hdr, seq->upper);
801 /* Build message as chain of buffers */
802 skb_queue_head_init(&pkts);
803 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
805 /* Send message if build was successful */
806 if (unlikely(rc == dlen))
807 rc = tipc_mcast_xmit(net, &pkts, method, &dsts,
808 &tsk->cong_link_cnt);
810 tipc_nlist_purge(&dsts);
812 return rc ? rc : dlen;
816 * tipc_send_group_msg - send a message to a member in the group
817 * @net: network namespace
818 * @m: message to send
820 * @dnode: destination node
821 * @dport: destination port
822 * @dlen: total length of message data
824 static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
825 struct msghdr *m, struct tipc_member *mb,
826 u32 dnode, u32 dport, int dlen)
828 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
829 struct tipc_mc_method *method = &tsk->mc_method;
830 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
831 struct tipc_msg *hdr = &tsk->phdr;
832 struct sk_buff_head pkts;
835 /* Complete message header */
836 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
837 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
838 msg_set_destport(hdr, dport);
839 msg_set_destnode(hdr, dnode);
840 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
842 /* Build message as chain of buffers */
843 skb_queue_head_init(&pkts);
844 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
845 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
846 if (unlikely(rc != dlen))
850 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
851 if (unlikely(rc == -ELINKCONG)) {
852 tipc_dest_push(&tsk->cong_links, dnode, 0);
853 tsk->cong_link_cnt++;
856 /* Update send window */
857 tipc_group_update_member(mb, blks);
859 /* A broadcast sent within next EXPIRE period must follow same path */
860 method->rcast = true;
861 method->mandatory = true;
866 * tipc_send_group_unicast - send message to a member in the group
867 * @sock: socket structure
868 * @m: message to send
869 * @dlen: total length of message data
870 * @timeout: timeout to wait for wakeup
872 * Called from function tipc_sendmsg(), which has done all sanity checks
873 * Returns the number of bytes sent on success, or errno
875 static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
876 int dlen, long timeout)
878 struct sock *sk = sock->sk;
879 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
880 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
881 struct tipc_sock *tsk = tipc_sk(sk);
882 struct tipc_group *grp = tsk->group;
883 struct net *net = sock_net(sk);
884 struct tipc_member *mb = NULL;
888 node = dest->addr.id.node;
889 port = dest->addr.id.ref;
891 return -EHOSTUNREACH;
893 /* Block or return if destination link or member is congested */
894 rc = tipc_wait_for_cond(sock, &timeout,
895 !tipc_dest_find(&tsk->cong_links, node, 0) &&
896 !tipc_group_cong(grp, node, port, blks, &mb));
901 return -EHOSTUNREACH;
903 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
905 return rc ? rc : dlen;
909 * tipc_send_group_anycast - send message to any member with given identity
910 * @sock: socket structure
911 * @m: message to send
912 * @dlen: total length of message data
913 * @timeout: timeout to wait for wakeup
915 * Called from function tipc_sendmsg(), which has done all sanity checks
916 * Returns the number of bytes sent on success, or errno
918 static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
919 int dlen, long timeout)
921 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
922 struct sock *sk = sock->sk;
923 struct tipc_sock *tsk = tipc_sk(sk);
924 struct list_head *cong_links = &tsk->cong_links;
925 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
926 struct tipc_group *grp = tsk->group;
927 struct tipc_msg *hdr = &tsk->phdr;
928 struct tipc_member *first = NULL;
929 struct tipc_member *mbr = NULL;
930 struct net *net = sock_net(sk);
931 u32 node, port, exclude;
932 struct list_head dsts;
933 u32 type, inst, scope;
938 INIT_LIST_HEAD(&dsts);
940 type = msg_nametype(hdr);
941 inst = dest->addr.name.name.instance;
942 scope = msg_lookup_scope(hdr);
943 exclude = tipc_group_exclude(grp);
945 while (++lookups < 4) {
948 /* Look for a non-congested destination member, if any */
950 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
951 &dstcnt, exclude, false))
952 return -EHOSTUNREACH;
953 tipc_dest_pop(&dsts, &node, &port);
954 cong = tipc_group_cong(grp, node, port, blks, &mbr);
963 /* Start over if destination was not in member list */
967 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
970 /* Block or return if destination link or member is congested */
971 rc = tipc_wait_for_cond(sock, &timeout,
972 !tipc_dest_find(cong_links, node, 0) &&
973 !tipc_group_cong(grp, node, port,
978 /* Send, unless destination disappeared while waiting */
983 if (unlikely(lookups >= 4))
984 return -EHOSTUNREACH;
986 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
988 return rc ? rc : dlen;
992 * tipc_send_group_bcast - send message to all members in communication group
993 * @sk: socket structure
994 * @m: message to send
995 * @dlen: total length of message data
996 * @timeout: timeout to wait for wakeup
998 * Called from function tipc_sendmsg(), which has done all sanity checks
999 * Returns the number of bytes sent on success, or errno
1001 static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1002 int dlen, long timeout)
1004 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1005 struct sock *sk = sock->sk;
1006 struct net *net = sock_net(sk);
1007 struct tipc_sock *tsk = tipc_sk(sk);
1008 struct tipc_group *grp = tsk->group;
1009 struct tipc_nlist *dsts = tipc_group_dests(grp);
1010 struct tipc_mc_method *method = &tsk->mc_method;
1011 bool ack = method->mandatory && method->rcast;
1012 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1013 struct tipc_msg *hdr = &tsk->phdr;
1014 int mtu = tipc_bcast_get_mtu(net);
1015 struct sk_buff_head pkts;
1016 int rc = -EHOSTUNREACH;
1018 if (!dsts->local && !dsts->remote)
1019 return -EHOSTUNREACH;
1021 /* Block or return if any destination link or member is congested */
1022 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt &&
1023 !tipc_group_bc_cong(grp, blks));
1027 /* Complete message header */
1029 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1030 msg_set_nameinst(hdr, dest->addr.name.name.instance);
1032 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1033 msg_set_nameinst(hdr, 0);
1035 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1036 msg_set_destport(hdr, 0);
1037 msg_set_destnode(hdr, 0);
1038 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(grp));
1040 /* Avoid getting stuck with repeated forced replicasts */
1041 msg_set_grp_bc_ack_req(hdr, ack);
1043 /* Build message as chain of buffers */
1044 skb_queue_head_init(&pkts);
1045 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1046 if (unlikely(rc != dlen))
1050 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1054 /* Update broadcast sequence number and send windows */
1055 tipc_group_update_bc_members(tsk->group, blks, ack);
1057 /* Broadcast link is now free to choose method for next broadcast */
1058 method->mandatory = false;
1059 method->expires = jiffies;
1065 * tipc_send_group_mcast - send message to all members with given identity
1066 * @sock: socket structure
1067 * @m: message to send
1068 * @dlen: total length of message data
1069 * @timeout: timeout to wait for wakeup
1071 * Called from function tipc_sendmsg(), which has done all sanity checks
1072 * Returns the number of bytes sent on success, or errno
1074 static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1075 int dlen, long timeout)
1077 struct sock *sk = sock->sk;
1078 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1079 struct tipc_sock *tsk = tipc_sk(sk);
1080 struct tipc_group *grp = tsk->group;
1081 struct tipc_msg *hdr = &tsk->phdr;
1082 struct net *net = sock_net(sk);
1083 u32 type, inst, scope, exclude;
1084 struct list_head dsts;
1087 INIT_LIST_HEAD(&dsts);
1089 type = msg_nametype(hdr);
1090 inst = dest->addr.name.name.instance;
1091 scope = msg_lookup_scope(hdr);
1092 exclude = tipc_group_exclude(grp);
1094 if (!tipc_nametbl_lookup(net, type, inst, scope, &dsts,
1095 &dstcnt, exclude, true))
1096 return -EHOSTUNREACH;
1099 tipc_dest_pop(&dsts, &dest->addr.id.node, &dest->addr.id.ref);
1100 return tipc_send_group_unicast(sock, m, dlen, timeout);
1103 tipc_dest_list_purge(&dsts);
1104 return tipc_send_group_bcast(sock, m, dlen, timeout);
1108 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1109 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1110 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1112 * Multi-threaded: parallel calls with reference to same queues may occur
1114 void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1115 struct sk_buff_head *inputq)
1117 u32 self = tipc_own_addr(net);
1118 u32 type, lower, upper, scope;
1119 struct sk_buff *skb, *_skb;
1120 u32 portid, oport, onode;
1121 struct sk_buff_head tmpq;
1122 struct list_head dports;
1123 struct tipc_msg *hdr;
1124 int user, mtyp, hlen;
1127 __skb_queue_head_init(&tmpq);
1128 INIT_LIST_HEAD(&dports);
1130 skb = tipc_skb_peek(arrvq, &inputq->lock);
1131 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1133 user = msg_user(hdr);
1134 mtyp = msg_type(hdr);
1135 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1136 oport = msg_origport(hdr);
1137 onode = msg_orignode(hdr);
1138 type = msg_nametype(hdr);
1140 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1141 spin_lock_bh(&inputq->lock);
1142 if (skb_peek(arrvq) == skb) {
1143 __skb_dequeue(arrvq);
1144 __skb_queue_tail(inputq, skb);
1147 spin_unlock_bh(&inputq->lock);
1151 /* Group messages require exact scope match */
1152 if (msg_in_group(hdr)) {
1155 scope = msg_lookup_scope(hdr);
1158 /* TIPC_NODE_SCOPE means "any scope" in this context */
1160 scope = TIPC_NODE_SCOPE;
1162 scope = TIPC_CLUSTER_SCOPE;
1164 lower = msg_namelower(hdr);
1165 upper = msg_nameupper(hdr);
1168 /* Create destination port list: */
1169 tipc_nametbl_mc_lookup(net, type, lower, upper,
1170 scope, exact, &dports);
1172 /* Clone message per destination */
1173 while (tipc_dest_pop(&dports, NULL, &portid)) {
1174 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1176 msg_set_destport(buf_msg(_skb), portid);
1177 __skb_queue_tail(&tmpq, _skb);
1180 pr_warn("Failed to clone mcast rcv buffer\n");
1182 /* Append to inputq if not already done by other thread */
1183 spin_lock_bh(&inputq->lock);
1184 if (skb_peek(arrvq) == skb) {
1185 skb_queue_splice_tail_init(&tmpq, inputq);
1186 kfree_skb(__skb_dequeue(arrvq));
1188 spin_unlock_bh(&inputq->lock);
1189 __skb_queue_purge(&tmpq);
1192 tipc_sk_rcv(net, inputq);
1196 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1197 * @tsk: receiving socket
1198 * @skb: pointer to message buffer.
1200 static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1201 struct sk_buff_head *xmitq)
1203 struct tipc_msg *hdr = buf_msg(skb);
1204 u32 onode = tsk_own_node(tsk);
1205 struct sock *sk = &tsk->sk;
1206 int mtyp = msg_type(hdr);
1209 /* Ignore if connection cannot be validated: */
1210 if (!tsk_peer_msg(tsk, hdr))
1213 if (unlikely(msg_errcode(hdr))) {
1214 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1215 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1216 tsk_peer_port(tsk));
1217 sk->sk_state_change(sk);
1221 tsk->probe_unacked = false;
1223 if (mtyp == CONN_PROBE) {
1224 msg_set_type(hdr, CONN_PROBE_REPLY);
1225 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1226 __skb_queue_tail(xmitq, skb);
1228 } else if (mtyp == CONN_ACK) {
1229 conn_cong = tsk_conn_cong(tsk);
1230 tsk->snt_unacked -= msg_conn_ack(hdr);
1231 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1232 tsk->snd_win = msg_adv_win(hdr);
1234 sk->sk_write_space(sk);
1235 } else if (mtyp != CONN_PROBE_REPLY) {
1236 pr_warn("Received unknown CONN_PROTO msg\n");
1243 * tipc_sendmsg - send message in connectionless manner
1244 * @sock: socket structure
1245 * @m: message to send
1246 * @dsz: amount of user data to be sent
1248 * Message must have an destination specified explicitly.
1249 * Used for SOCK_RDM and SOCK_DGRAM messages,
1250 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1251 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1253 * Returns the number of bytes sent on success, or errno otherwise
1255 static int tipc_sendmsg(struct socket *sock,
1256 struct msghdr *m, size_t dsz)
1258 struct sock *sk = sock->sk;
1262 ret = __tipc_sendmsg(sock, m, dsz);
1268 static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1270 struct sock *sk = sock->sk;
1271 struct net *net = sock_net(sk);
1272 struct tipc_sock *tsk = tipc_sk(sk);
1273 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1274 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1275 struct list_head *clinks = &tsk->cong_links;
1276 bool syn = !tipc_sk_type_connectionless(sk);
1277 struct tipc_group *grp = tsk->group;
1278 struct tipc_msg *hdr = &tsk->phdr;
1279 struct tipc_name_seq *seq;
1280 struct sk_buff_head pkts;
1281 u32 dport, dnode = 0;
1285 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1289 if (unlikely(m->msg_namelen < sizeof(*dest)))
1291 if (unlikely(dest->family != AF_TIPC))
1297 return tipc_send_group_bcast(sock, m, dlen, timeout);
1298 if (dest->addrtype == TIPC_ADDR_NAME)
1299 return tipc_send_group_anycast(sock, m, dlen, timeout);
1300 if (dest->addrtype == TIPC_ADDR_ID)
1301 return tipc_send_group_unicast(sock, m, dlen, timeout);
1302 if (dest->addrtype == TIPC_ADDR_MCAST)
1303 return tipc_send_group_mcast(sock, m, dlen, timeout);
1307 if (unlikely(!dest)) {
1309 if (!syn || dest->family != AF_TIPC)
1310 return -EDESTADDRREQ;
1313 if (unlikely(syn)) {
1314 if (sk->sk_state == TIPC_LISTEN)
1316 if (sk->sk_state != TIPC_OPEN)
1320 if (dest->addrtype == TIPC_ADDR_NAME) {
1321 tsk->conn_type = dest->addr.name.name.type;
1322 tsk->conn_instance = dest->addr.name.name.instance;
1326 seq = &dest->addr.nameseq;
1327 if (dest->addrtype == TIPC_ADDR_MCAST)
1328 return tipc_sendmcast(sock, seq, m, dlen, timeout);
1330 if (dest->addrtype == TIPC_ADDR_NAME) {
1331 type = dest->addr.name.name.type;
1332 inst = dest->addr.name.name.instance;
1333 dnode = dest->addr.name.domain;
1334 msg_set_type(hdr, TIPC_NAMED_MSG);
1335 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1336 msg_set_nametype(hdr, type);
1337 msg_set_nameinst(hdr, inst);
1338 msg_set_lookup_scope(hdr, tipc_node2scope(dnode));
1339 dport = tipc_nametbl_translate(net, type, inst, &dnode);
1340 msg_set_destnode(hdr, dnode);
1341 msg_set_destport(hdr, dport);
1342 if (unlikely(!dport && !dnode))
1343 return -EHOSTUNREACH;
1344 } else if (dest->addrtype == TIPC_ADDR_ID) {
1345 dnode = dest->addr.id.node;
1346 msg_set_type(hdr, TIPC_DIRECT_MSG);
1347 msg_set_lookup_scope(hdr, 0);
1348 msg_set_destnode(hdr, dnode);
1349 msg_set_destport(hdr, dest->addr.id.ref);
1350 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1355 /* Block or return if destination link is congested */
1356 rc = tipc_wait_for_cond(sock, &timeout,
1357 !tipc_dest_find(clinks, dnode, 0));
1361 skb_queue_head_init(&pkts);
1362 mtu = tipc_node_get_mtu(net, dnode, tsk->portid);
1363 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1364 if (unlikely(rc != dlen))
1367 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1368 if (unlikely(rc == -ELINKCONG)) {
1369 tipc_dest_push(clinks, dnode, 0);
1370 tsk->cong_link_cnt++;
1374 if (unlikely(syn && !rc))
1375 tipc_set_sk_state(sk, TIPC_CONNECTING);
1377 return rc ? rc : dlen;
1381 * tipc_sendstream - send stream-oriented data
1382 * @sock: socket structure
1384 * @dsz: total length of data to be transmitted
1386 * Used for SOCK_STREAM data.
1388 * Returns the number of bytes sent on success (or partial success),
1389 * or errno if no data sent
1391 static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1393 struct sock *sk = sock->sk;
1397 ret = __tipc_sendstream(sock, m, dsz);
1403 static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1405 struct sock *sk = sock->sk;
1406 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1407 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1408 struct tipc_sock *tsk = tipc_sk(sk);
1409 struct tipc_msg *hdr = &tsk->phdr;
1410 struct net *net = sock_net(sk);
1411 struct sk_buff_head pkts;
1412 u32 dnode = tsk_peer_node(tsk);
1416 skb_queue_head_init(&pkts);
1418 if (unlikely(dlen > INT_MAX))
1421 /* Handle implicit connection setup */
1422 if (unlikely(dest)) {
1423 rc = __tipc_sendmsg(sock, m, dlen);
1424 if (dlen && (dlen == rc))
1425 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1430 rc = tipc_wait_for_cond(sock, &timeout,
1431 (!tsk->cong_link_cnt &&
1432 !tsk_conn_cong(tsk) &&
1433 tipc_sk_connected(sk)));
1437 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1438 rc = tipc_msg_build(hdr, m, sent, send, tsk->max_pkt, &pkts);
1439 if (unlikely(rc != send))
1442 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
1443 if (unlikely(rc == -ELINKCONG)) {
1444 tsk->cong_link_cnt = 1;
1448 tsk->snt_unacked += tsk_inc(tsk, send + MIN_H_SIZE);
1451 } while (sent < dlen && !rc);
1453 return sent ? sent : rc;
1457 * tipc_send_packet - send a connection-oriented message
1458 * @sock: socket structure
1459 * @m: message to send
1460 * @dsz: length of data to be transmitted
1462 * Used for SOCK_SEQPACKET messages.
1464 * Returns the number of bytes sent on success, or errno otherwise
1466 static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1468 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1471 return tipc_sendstream(sock, m, dsz);
1474 /* tipc_sk_finish_conn - complete the setup of a connection
1476 static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1479 struct sock *sk = &tsk->sk;
1480 struct net *net = sock_net(sk);
1481 struct tipc_msg *msg = &tsk->phdr;
1483 msg_set_destnode(msg, peer_node);
1484 msg_set_destport(msg, peer_port);
1485 msg_set_type(msg, TIPC_CONN_MSG);
1486 msg_set_lookup_scope(msg, 0);
1487 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1489 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1490 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1491 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1492 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid);
1493 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1494 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1497 /* Fall back to message based flow control */
1498 tsk->rcv_win = FLOWCTL_MSG_WIN;
1499 tsk->snd_win = FLOWCTL_MSG_WIN;
1503 * tipc_sk_set_orig_addr - capture sender's address for received message
1504 * @m: descriptor for message info
1505 * @hdr: received message header
1507 * Note: Address is not captured if not requested by receiver.
1509 static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1511 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1512 struct tipc_msg *hdr = buf_msg(skb);
1517 srcaddr->sock.family = AF_TIPC;
1518 srcaddr->sock.addrtype = TIPC_ADDR_ID;
1519 srcaddr->sock.scope = 0;
1520 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1521 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1522 srcaddr->sock.addr.name.domain = 0;
1523 m->msg_namelen = sizeof(struct sockaddr_tipc);
1525 if (!msg_in_group(hdr))
1528 /* Group message users may also want to know sending member's id */
1529 srcaddr->member.family = AF_TIPC;
1530 srcaddr->member.addrtype = TIPC_ADDR_NAME;
1531 srcaddr->member.scope = 0;
1532 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1533 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1534 srcaddr->member.addr.name.domain = 0;
1535 m->msg_namelen = sizeof(*srcaddr);
1539 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1540 * @m: descriptor for message info
1541 * @msg: received message header
1542 * @tsk: TIPC port associated with message
1544 * Note: Ancillary data is not captured if not requested by receiver.
1546 * Returns 0 if successful, otherwise errno
1548 static int tipc_sk_anc_data_recv(struct msghdr *m, struct tipc_msg *msg,
1549 struct tipc_sock *tsk)
1557 if (likely(m->msg_controllen == 0))
1560 /* Optionally capture errored message object(s) */
1561 err = msg ? msg_errcode(msg) : 0;
1562 if (unlikely(err)) {
1564 anc_data[1] = msg_data_sz(msg);
1565 res = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, anc_data);
1569 res = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, anc_data[1],
1576 /* Optionally capture message destination object */
1577 dest_type = msg ? msg_type(msg) : TIPC_DIRECT_MSG;
1578 switch (dest_type) {
1579 case TIPC_NAMED_MSG:
1581 anc_data[0] = msg_nametype(msg);
1582 anc_data[1] = msg_namelower(msg);
1583 anc_data[2] = msg_namelower(msg);
1585 case TIPC_MCAST_MSG:
1587 anc_data[0] = msg_nametype(msg);
1588 anc_data[1] = msg_namelower(msg);
1589 anc_data[2] = msg_nameupper(msg);
1592 has_name = (tsk->conn_type != 0);
1593 anc_data[0] = tsk->conn_type;
1594 anc_data[1] = tsk->conn_instance;
1595 anc_data[2] = tsk->conn_instance;
1601 res = put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, anc_data);
1609 static void tipc_sk_send_ack(struct tipc_sock *tsk)
1611 struct sock *sk = &tsk->sk;
1612 struct net *net = sock_net(sk);
1613 struct sk_buff *skb = NULL;
1614 struct tipc_msg *msg;
1615 u32 peer_port = tsk_peer_port(tsk);
1616 u32 dnode = tsk_peer_node(tsk);
1618 if (!tipc_sk_connected(sk))
1620 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1621 dnode, tsk_own_node(tsk), peer_port,
1622 tsk->portid, TIPC_OK);
1626 msg_set_conn_ack(msg, tsk->rcv_unacked);
1627 tsk->rcv_unacked = 0;
1629 /* Adjust to and advertize the correct window limit */
1630 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1631 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1632 msg_set_adv_win(msg, tsk->rcv_win);
1634 tipc_node_xmit_skb(net, skb, dnode, msg_link_selector(msg));
1637 static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1639 struct sock *sk = sock->sk;
1641 long timeo = *timeop;
1642 int err = sock_error(sk);
1648 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1649 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1650 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1655 timeo = schedule_timeout(timeo);
1659 if (!skb_queue_empty(&sk->sk_receive_queue))
1664 err = sock_intr_errno(timeo);
1665 if (signal_pending(current))
1668 err = sock_error(sk);
1672 finish_wait(sk_sleep(sk), &wait);
1678 * tipc_recvmsg - receive packet-oriented message
1679 * @m: descriptor for message info
1680 * @buflen: length of user buffer area
1681 * @flags: receive flags
1683 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1684 * If the complete message doesn't fit in user area, truncate it.
1686 * Returns size of returned message data, errno otherwise
1688 static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1689 size_t buflen, int flags)
1691 struct sock *sk = sock->sk;
1692 bool connected = !tipc_sk_type_connectionless(sk);
1693 struct tipc_sock *tsk = tipc_sk(sk);
1694 int rc, err, hlen, dlen, copy;
1695 struct sk_buff_head xmitq;
1696 struct tipc_msg *hdr;
1697 struct sk_buff *skb;
1701 /* Catch invalid receive requests */
1702 if (unlikely(!buflen))
1706 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1710 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1712 /* Step rcv queue to first msg with data or error; wait if necessary */
1714 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1717 skb = skb_peek(&sk->sk_receive_queue);
1719 dlen = msg_data_sz(hdr);
1720 hlen = msg_hdr_sz(hdr);
1721 err = msg_errcode(hdr);
1722 grp_evt = msg_is_grp_evt(hdr);
1723 if (likely(dlen || err))
1725 tsk_advance_rx_queue(sk);
1728 /* Collect msg meta data, including error code and rejected data */
1729 tipc_sk_set_orig_addr(m, skb);
1730 rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1734 /* Capture data if non-error msg, otherwise just set return value */
1736 copy = min_t(int, dlen, buflen);
1737 if (unlikely(copy != dlen))
1738 m->msg_flags |= MSG_TRUNC;
1739 rc = skb_copy_datagram_msg(skb, hlen, m, copy);
1743 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control)
1749 /* Mark message as group event if applicable */
1750 if (unlikely(grp_evt)) {
1751 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1752 m->msg_flags |= MSG_EOR;
1753 m->msg_flags |= MSG_OOB;
1757 /* Caption of data or error code/rejected data was successful */
1758 if (unlikely(flags & MSG_PEEK))
1761 /* Send group flow control advertisement when applicable */
1762 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1763 skb_queue_head_init(&xmitq);
1764 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1765 msg_orignode(hdr), msg_origport(hdr),
1767 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1770 tsk_advance_rx_queue(sk);
1772 if (likely(!connected))
1775 /* Send connection flow control advertisement when applicable */
1776 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1777 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1778 tipc_sk_send_ack(tsk);
1781 return rc ? rc : copy;
1785 * tipc_recvstream - receive stream-oriented data
1786 * @m: descriptor for message info
1787 * @buflen: total size of user buffer area
1788 * @flags: receive flags
1790 * Used for SOCK_STREAM messages only. If not enough data is available
1791 * will optionally wait for more; never truncates data.
1793 * Returns size of returned message data, errno otherwise
1795 static int tipc_recvstream(struct socket *sock, struct msghdr *m,
1796 size_t buflen, int flags)
1798 struct sock *sk = sock->sk;
1799 struct tipc_sock *tsk = tipc_sk(sk);
1800 struct sk_buff *skb;
1801 struct tipc_msg *hdr;
1802 struct tipc_skb_cb *skb_cb;
1803 bool peek = flags & MSG_PEEK;
1804 int offset, required, copy, copied = 0;
1805 int hlen, dlen, err, rc;
1808 /* Catch invalid receive attempts */
1809 if (unlikely(!buflen))
1814 if (unlikely(sk->sk_state == TIPC_OPEN)) {
1818 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
1819 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1822 /* Look at first msg in receive queue; wait if necessary */
1823 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1826 skb = skb_peek(&sk->sk_receive_queue);
1827 skb_cb = TIPC_SKB_CB(skb);
1829 dlen = msg_data_sz(hdr);
1830 hlen = msg_hdr_sz(hdr);
1831 err = msg_errcode(hdr);
1833 /* Discard any empty non-errored (SYN-) message */
1834 if (unlikely(!dlen && !err)) {
1835 tsk_advance_rx_queue(sk);
1839 /* Collect msg meta data, incl. error code and rejected data */
1841 tipc_sk_set_orig_addr(m, skb);
1842 rc = tipc_sk_anc_data_recv(m, hdr, tsk);
1847 /* Copy data if msg ok, otherwise return error/partial data */
1849 offset = skb_cb->bytes_read;
1850 copy = min_t(int, dlen - offset, buflen - copied);
1851 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1856 if (unlikely(offset < dlen)) {
1858 skb_cb->bytes_read = offset;
1863 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
1872 tsk_advance_rx_queue(sk);
1874 /* Send connection flow control advertisement when applicable */
1875 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1876 if (unlikely(tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE))
1877 tipc_sk_send_ack(tsk);
1879 /* Exit if all requested data or FIN/error received */
1880 if (copied == buflen || err)
1883 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
1886 return copied ? copied : rc;
1890 * tipc_write_space - wake up thread if port congestion is released
1893 static void tipc_write_space(struct sock *sk)
1895 struct socket_wq *wq;
1898 wq = rcu_dereference(sk->sk_wq);
1899 if (skwq_has_sleeper(wq))
1900 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
1901 EPOLLWRNORM | EPOLLWRBAND);
1906 * tipc_data_ready - wake up threads to indicate messages have been received
1908 * @len: the length of messages
1910 static void tipc_data_ready(struct sock *sk)
1912 struct socket_wq *wq;
1915 wq = rcu_dereference(sk->sk_wq);
1916 if (skwq_has_sleeper(wq))
1917 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
1918 EPOLLRDNORM | EPOLLRDBAND);
1922 static void tipc_sock_destruct(struct sock *sk)
1924 __skb_queue_purge(&sk->sk_receive_queue);
1927 static void tipc_sk_proto_rcv(struct sock *sk,
1928 struct sk_buff_head *inputq,
1929 struct sk_buff_head *xmitq)
1931 struct sk_buff *skb = __skb_dequeue(inputq);
1932 struct tipc_sock *tsk = tipc_sk(sk);
1933 struct tipc_msg *hdr = buf_msg(skb);
1934 struct tipc_group *grp = tsk->group;
1935 bool wakeup = false;
1937 switch (msg_user(hdr)) {
1939 tipc_sk_conn_proto_rcv(tsk, skb, xmitq);
1942 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
1943 tsk->cong_link_cnt--;
1946 case GROUP_PROTOCOL:
1947 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
1950 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
1951 hdr, inputq, xmitq);
1958 sk->sk_write_space(sk);
1964 * tipc_filter_connect - Handle incoming message for a connection-based socket
1966 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1968 * Returns true if everything ok, false otherwise
1970 static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb)
1972 struct sock *sk = &tsk->sk;
1973 struct net *net = sock_net(sk);
1974 struct tipc_msg *hdr = buf_msg(skb);
1975 u32 pport = msg_origport(hdr);
1976 u32 pnode = msg_orignode(hdr);
1978 if (unlikely(msg_mcast(hdr)))
1981 switch (sk->sk_state) {
1982 case TIPC_CONNECTING:
1983 /* Accept only ACK or NACK message */
1984 if (unlikely(!msg_connected(hdr))) {
1985 if (pport != tsk_peer_port(tsk) ||
1986 pnode != tsk_peer_node(tsk))
1989 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1990 sk->sk_err = ECONNREFUSED;
1991 sk->sk_state_change(sk);
1995 if (unlikely(msg_errcode(hdr))) {
1996 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1997 sk->sk_err = ECONNREFUSED;
1998 sk->sk_state_change(sk);
2002 if (unlikely(!msg_isdata(hdr))) {
2003 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2004 sk->sk_err = EINVAL;
2005 sk->sk_state_change(sk);
2009 tipc_sk_finish_conn(tsk, msg_origport(hdr), msg_orignode(hdr));
2010 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2012 /* If 'ACK+' message, add to socket receive queue */
2013 if (msg_data_sz(hdr))
2016 /* If empty 'ACK-' message, wake up sleeping connect() */
2017 sk->sk_data_ready(sk);
2019 /* 'ACK-' message is neither accepted nor rejected: */
2020 msg_set_dest_droppable(hdr, 1);
2024 case TIPC_DISCONNECTING:
2027 /* Accept only SYN message */
2028 if (!msg_connected(hdr) && !(msg_errcode(hdr)))
2031 case TIPC_ESTABLISHED:
2032 /* Accept only connection-based messages sent by peer */
2033 if (unlikely(!tsk_peer_msg(tsk, hdr)))
2036 if (unlikely(msg_errcode(hdr))) {
2037 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2038 /* Let timer expire on it's own */
2039 tipc_node_remove_conn(net, tsk_peer_node(tsk),
2041 sk->sk_state_change(sk);
2045 pr_err("Unknown sk_state %u\n", sk->sk_state);
2052 * rcvbuf_limit - get proper overload limit of socket receive queue
2056 * For connection oriented messages, irrespective of importance,
2057 * default queue limit is 2 MB.
2059 * For connectionless messages, queue limits are based on message
2060 * importance as follows:
2062 * TIPC_LOW_IMPORTANCE (2 MB)
2063 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2064 * TIPC_HIGH_IMPORTANCE (8 MB)
2065 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2067 * Returns overload limit according to corresponding message importance
2069 static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2071 struct tipc_sock *tsk = tipc_sk(sk);
2072 struct tipc_msg *hdr = buf_msg(skb);
2074 if (unlikely(msg_in_group(hdr)))
2075 return sk->sk_rcvbuf;
2077 if (unlikely(!msg_connected(hdr)))
2078 return sk->sk_rcvbuf << msg_importance(hdr);
2080 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2081 return sk->sk_rcvbuf;
2083 return FLOWCTL_MSG_LIM;
2087 * tipc_sk_filter_rcv - validate incoming message
2089 * @skb: pointer to message.
2091 * Enqueues message on receive queue if acceptable; optionally handles
2092 * disconnect indication for a connected socket.
2094 * Called with socket lock already taken
2097 static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2098 struct sk_buff_head *xmitq)
2100 bool sk_conn = !tipc_sk_type_connectionless(sk);
2101 struct tipc_sock *tsk = tipc_sk(sk);
2102 struct tipc_group *grp = tsk->group;
2103 struct tipc_msg *hdr = buf_msg(skb);
2104 struct net *net = sock_net(sk);
2105 struct sk_buff_head inputq;
2106 int limit, err = TIPC_OK;
2108 TIPC_SKB_CB(skb)->bytes_read = 0;
2109 __skb_queue_head_init(&inputq);
2110 __skb_queue_tail(&inputq, skb);
2112 if (unlikely(!msg_isdata(hdr)))
2113 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2116 tipc_group_filter_msg(grp, &inputq, xmitq);
2118 /* Validate and add to receive buffer if there is space */
2119 while ((skb = __skb_dequeue(&inputq))) {
2121 limit = rcvbuf_limit(sk, skb);
2122 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb)) ||
2123 (!sk_conn && msg_connected(hdr)) ||
2124 (!grp && msg_in_group(hdr)))
2125 err = TIPC_ERR_NO_PORT;
2126 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2127 atomic_inc(&sk->sk_drops);
2128 err = TIPC_ERR_OVERLOAD;
2131 if (unlikely(err)) {
2132 tipc_skb_reject(net, err, skb, xmitq);
2136 __skb_queue_tail(&sk->sk_receive_queue, skb);
2137 skb_set_owner_r(skb, sk);
2138 sk->sk_data_ready(sk);
2143 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2147 * Caller must hold socket lock
2149 static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2151 unsigned int before = sk_rmem_alloc_get(sk);
2152 struct sk_buff_head xmitq;
2155 __skb_queue_head_init(&xmitq);
2157 tipc_sk_filter_rcv(sk, skb, &xmitq);
2158 added = sk_rmem_alloc_get(sk) - before;
2159 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2161 /* Send pending response/rejected messages, if any */
2162 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2167 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2168 * inputq and try adding them to socket or backlog queue
2169 * @inputq: list of incoming buffers with potentially different destinations
2170 * @sk: socket where the buffers should be enqueued
2171 * @dport: port number for the socket
2173 * Caller must hold socket lock
2175 static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2176 u32 dport, struct sk_buff_head *xmitq)
2178 unsigned long time_limit = jiffies + 2;
2179 struct sk_buff *skb;
2184 while (skb_queue_len(inputq)) {
2185 if (unlikely(time_after_eq(jiffies, time_limit)))
2188 skb = tipc_skb_dequeue(inputq, dport);
2192 /* Add message directly to receive queue if possible */
2193 if (!sock_owned_by_user(sk)) {
2194 tipc_sk_filter_rcv(sk, skb, xmitq);
2198 /* Try backlog, compensating for double-counted bytes */
2199 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2200 if (!sk->sk_backlog.len)
2201 atomic_set(dcnt, 0);
2202 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2203 if (likely(!sk_add_backlog(sk, skb, lim)))
2206 /* Overload => reject message back to sender */
2207 onode = tipc_own_addr(sock_net(sk));
2208 atomic_inc(&sk->sk_drops);
2209 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD))
2210 __skb_queue_tail(xmitq, skb);
2216 * tipc_sk_rcv - handle a chain of incoming buffers
2217 * @inputq: buffer list containing the buffers
2218 * Consumes all buffers in list until inputq is empty
2219 * Note: may be called in multiple threads referring to the same queue
2221 void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2223 struct sk_buff_head xmitq;
2224 u32 dnode, dport = 0;
2226 struct tipc_sock *tsk;
2228 struct sk_buff *skb;
2230 __skb_queue_head_init(&xmitq);
2231 while (skb_queue_len(inputq)) {
2232 dport = tipc_skb_peek_port(inputq, dport);
2233 tsk = tipc_sk_lookup(net, dport);
2237 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2238 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2239 spin_unlock_bh(&sk->sk_lock.slock);
2241 /* Send pending response/rejected messages, if any */
2242 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2246 /* No destination socket => dequeue skb if still there */
2247 skb = tipc_skb_dequeue(inputq, dport);
2251 /* Try secondary lookup if unresolved named message */
2252 err = TIPC_ERR_NO_PORT;
2253 if (tipc_msg_lookup_dest(net, skb, &err))
2256 /* Prepare for message rejection */
2257 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2260 dnode = msg_destnode(buf_msg(skb));
2261 tipc_node_xmit_skb(net, skb, dnode, dport);
2265 static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2267 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2268 struct sock *sk = sock->sk;
2272 int err = sock_error(sk);
2277 if (signal_pending(current))
2278 return sock_intr_errno(*timeo_p);
2280 add_wait_queue(sk_sleep(sk), &wait);
2281 done = sk_wait_event(sk, timeo_p,
2282 sk->sk_state != TIPC_CONNECTING, &wait);
2283 remove_wait_queue(sk_sleep(sk), &wait);
2289 * tipc_connect - establish a connection to another TIPC port
2290 * @sock: socket structure
2291 * @dest: socket address for destination port
2292 * @destlen: size of socket address data structure
2293 * @flags: file-related flags associated with socket
2295 * Returns 0 on success, errno otherwise
2297 static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2298 int destlen, int flags)
2300 struct sock *sk = sock->sk;
2301 struct tipc_sock *tsk = tipc_sk(sk);
2302 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2303 struct msghdr m = {NULL,};
2304 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2308 if (destlen != sizeof(struct sockaddr_tipc))
2318 if (dst->family == AF_UNSPEC) {
2319 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2320 if (!tipc_sk_type_connectionless(sk))
2323 } else if (dst->family != AF_TIPC) {
2326 if (dst->addrtype != TIPC_ADDR_ID && dst->addrtype != TIPC_ADDR_NAME)
2331 /* DGRAM/RDM connect(), just save the destaddr */
2332 if (tipc_sk_type_connectionless(sk)) {
2333 memcpy(&tsk->peer, dest, destlen);
2337 previous = sk->sk_state;
2339 switch (sk->sk_state) {
2341 /* Send a 'SYN-' to destination */
2343 m.msg_namelen = destlen;
2345 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2346 * indicate send_msg() is never blocked.
2349 m.msg_flags = MSG_DONTWAIT;
2351 res = __tipc_sendmsg(sock, &m, 0);
2352 if ((res < 0) && (res != -EWOULDBLOCK))
2355 /* Just entered TIPC_CONNECTING state; the only
2356 * difference is that return value in non-blocking
2357 * case is EINPROGRESS, rather than EALREADY.
2361 case TIPC_CONNECTING:
2363 if (previous == TIPC_CONNECTING)
2367 timeout = msecs_to_jiffies(timeout);
2368 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2369 res = tipc_wait_for_connect(sock, &timeout);
2371 case TIPC_ESTABLISHED:
2384 * tipc_listen - allow socket to listen for incoming connections
2385 * @sock: socket structure
2388 * Returns 0 on success, errno otherwise
2390 static int tipc_listen(struct socket *sock, int len)
2392 struct sock *sk = sock->sk;
2396 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2402 static int tipc_wait_for_accept(struct socket *sock, long timeo)
2404 struct sock *sk = sock->sk;
2408 /* True wake-one mechanism for incoming connections: only
2409 * one process gets woken up, not the 'whole herd'.
2410 * Since we do not 'race & poll' for established sockets
2411 * anymore, the common case will execute the loop only once.
2414 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
2415 TASK_INTERRUPTIBLE);
2416 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2418 timeo = schedule_timeout(timeo);
2422 if (!skb_queue_empty(&sk->sk_receive_queue))
2427 err = sock_intr_errno(timeo);
2428 if (signal_pending(current))
2431 finish_wait(sk_sleep(sk), &wait);
2436 * tipc_accept - wait for connection request
2437 * @sock: listening socket
2438 * @newsock: new socket that is to be connected
2439 * @flags: file-related flags associated with socket
2441 * Returns 0 on success, errno otherwise
2443 static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2446 struct sock *new_sk, *sk = sock->sk;
2447 struct sk_buff *buf;
2448 struct tipc_sock *new_tsock;
2449 struct tipc_msg *msg;
2455 if (sk->sk_state != TIPC_LISTEN) {
2459 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2460 res = tipc_wait_for_accept(sock, timeo);
2464 buf = skb_peek(&sk->sk_receive_queue);
2466 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2469 security_sk_clone(sock->sk, new_sock->sk);
2471 new_sk = new_sock->sk;
2472 new_tsock = tipc_sk(new_sk);
2475 /* we lock on new_sk; but lockdep sees the lock on sk */
2476 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2479 * Reject any stray messages received by new socket
2480 * before the socket lock was taken (very, very unlikely)
2482 tsk_rej_rx_queue(new_sk);
2484 /* Connect new socket to it's peer */
2485 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2487 tsk_set_importance(new_tsock, msg_importance(msg));
2488 if (msg_named(msg)) {
2489 new_tsock->conn_type = msg_nametype(msg);
2490 new_tsock->conn_instance = msg_nameinst(msg);
2494 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2495 * Respond to 'SYN+' by queuing it on new socket.
2497 if (!msg_data_sz(msg)) {
2498 struct msghdr m = {NULL,};
2500 tsk_advance_rx_queue(sk);
2501 __tipc_sendstream(new_sock, &m, 0);
2503 __skb_dequeue(&sk->sk_receive_queue);
2504 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2505 skb_set_owner_r(buf, new_sk);
2507 release_sock(new_sk);
2514 * tipc_shutdown - shutdown socket connection
2515 * @sock: socket structure
2516 * @how: direction to close (must be SHUT_RDWR)
2518 * Terminates connection (if necessary), then purges socket's receive queue.
2520 * Returns 0 on success, errno otherwise
2522 static int tipc_shutdown(struct socket *sock, int how)
2524 struct sock *sk = sock->sk;
2527 if (how != SHUT_RDWR)
2532 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2533 sk->sk_shutdown = SEND_SHUTDOWN;
2535 if (sk->sk_state == TIPC_DISCONNECTING) {
2536 /* Discard any unreceived messages */
2537 __skb_queue_purge(&sk->sk_receive_queue);
2539 /* Wake up anyone sleeping in poll */
2540 sk->sk_state_change(sk);
2550 static void tipc_sk_timeout(struct timer_list *t)
2552 struct sock *sk = from_timer(sk, t, sk_timer);
2553 struct tipc_sock *tsk = tipc_sk(sk);
2554 u32 peer_port = tsk_peer_port(tsk);
2555 u32 peer_node = tsk_peer_node(tsk);
2556 u32 own_node = tsk_own_node(tsk);
2557 u32 own_port = tsk->portid;
2558 struct net *net = sock_net(sk);
2559 struct sk_buff *skb = NULL;
2562 if (!tipc_sk_connected(sk))
2565 /* Try again later if socket is busy */
2566 if (sock_owned_by_user(sk)) {
2567 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2571 if (tsk->probe_unacked) {
2572 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2573 tipc_node_remove_conn(net, peer_node, peer_port);
2574 sk->sk_state_change(sk);
2577 /* Send new probe */
2578 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2579 peer_node, own_node, peer_port, own_port,
2581 tsk->probe_unacked = true;
2582 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2586 tipc_node_xmit_skb(net, skb, peer_node, own_port);
2590 static int tipc_sk_publish(struct tipc_sock *tsk, uint scope,
2591 struct tipc_name_seq const *seq)
2593 struct sock *sk = &tsk->sk;
2594 struct net *net = sock_net(sk);
2595 struct publication *publ;
2598 if (scope != TIPC_NODE_SCOPE)
2599 scope = TIPC_CLUSTER_SCOPE;
2601 if (tipc_sk_connected(sk))
2603 key = tsk->portid + tsk->pub_count + 1;
2604 if (key == tsk->portid)
2607 publ = tipc_nametbl_publish(net, seq->type, seq->lower, seq->upper,
2608 scope, tsk->portid, key);
2609 if (unlikely(!publ))
2612 list_add(&publ->binding_sock, &tsk->publications);
2618 static int tipc_sk_withdraw(struct tipc_sock *tsk, uint scope,
2619 struct tipc_name_seq const *seq)
2621 struct net *net = sock_net(&tsk->sk);
2622 struct publication *publ;
2623 struct publication *safe;
2626 if (scope != TIPC_NODE_SCOPE)
2627 scope = TIPC_CLUSTER_SCOPE;
2629 list_for_each_entry_safe(publ, safe, &tsk->publications, binding_sock) {
2631 if (publ->scope != scope)
2633 if (publ->type != seq->type)
2635 if (publ->lower != seq->lower)
2637 if (publ->upper != seq->upper)
2639 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2640 publ->upper, publ->key);
2644 tipc_nametbl_withdraw(net, publ->type, publ->lower,
2645 publ->upper, publ->key);
2648 if (list_empty(&tsk->publications))
2653 /* tipc_sk_reinit: set non-zero address in all existing sockets
2654 * when we go from standalone to network mode.
2656 void tipc_sk_reinit(struct net *net)
2658 struct tipc_net *tn = net_generic(net, tipc_net_id);
2659 struct rhashtable_iter iter;
2660 struct tipc_sock *tsk;
2661 struct tipc_msg *msg;
2663 rhashtable_walk_enter(&tn->sk_rht, &iter);
2666 rhashtable_walk_start(&iter);
2668 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2669 spin_lock_bh(&tsk->sk.sk_lock.slock);
2671 msg_set_prevnode(msg, tipc_own_addr(net));
2672 msg_set_orignode(msg, tipc_own_addr(net));
2673 spin_unlock_bh(&tsk->sk.sk_lock.slock);
2676 rhashtable_walk_stop(&iter);
2677 } while (tsk == ERR_PTR(-EAGAIN));
2680 static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2682 struct tipc_net *tn = net_generic(net, tipc_net_id);
2683 struct tipc_sock *tsk;
2686 tsk = rhashtable_lookup_fast(&tn->sk_rht, &portid, tsk_rht_params);
2688 sock_hold(&tsk->sk);
2694 static int tipc_sk_insert(struct tipc_sock *tsk)
2696 struct sock *sk = &tsk->sk;
2697 struct net *net = sock_net(sk);
2698 struct tipc_net *tn = net_generic(net, tipc_net_id);
2699 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
2700 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
2702 while (remaining--) {
2704 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
2705 portid = TIPC_MIN_PORT;
2706 tsk->portid = portid;
2707 sock_hold(&tsk->sk);
2708 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
2717 static void tipc_sk_remove(struct tipc_sock *tsk)
2719 struct sock *sk = &tsk->sk;
2720 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
2722 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
2723 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
2728 static const struct rhashtable_params tsk_rht_params = {
2730 .head_offset = offsetof(struct tipc_sock, node),
2731 .key_offset = offsetof(struct tipc_sock, portid),
2732 .key_len = sizeof(u32), /* portid */
2733 .max_size = 1048576,
2735 .automatic_shrinking = true,
2738 int tipc_sk_rht_init(struct net *net)
2740 struct tipc_net *tn = net_generic(net, tipc_net_id);
2742 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
2745 void tipc_sk_rht_destroy(struct net *net)
2747 struct tipc_net *tn = net_generic(net, tipc_net_id);
2749 /* Wait for socket readers to complete */
2752 rhashtable_destroy(&tn->sk_rht);
2755 static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
2757 struct net *net = sock_net(&tsk->sk);
2758 struct tipc_group *grp = tsk->group;
2759 struct tipc_msg *hdr = &tsk->phdr;
2760 struct tipc_name_seq seq;
2763 if (mreq->type < TIPC_RESERVED_TYPES)
2765 if (mreq->scope > TIPC_NODE_SCOPE)
2769 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
2773 msg_set_lookup_scope(hdr, mreq->scope);
2774 msg_set_nametype(hdr, mreq->type);
2775 msg_set_dest_droppable(hdr, true);
2776 seq.type = mreq->type;
2777 seq.lower = mreq->instance;
2778 seq.upper = seq.lower;
2779 tipc_nametbl_build_group(net, grp, mreq->type, mreq->scope);
2780 rc = tipc_sk_publish(tsk, mreq->scope, &seq);
2782 tipc_group_delete(net, grp);
2786 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2787 tsk->mc_method.rcast = true;
2788 tsk->mc_method.mandatory = true;
2789 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2793 static int tipc_sk_leave(struct tipc_sock *tsk)
2795 struct net *net = sock_net(&tsk->sk);
2796 struct tipc_group *grp = tsk->group;
2797 struct tipc_name_seq seq;
2802 tipc_group_self(grp, &seq, &scope);
2803 tipc_group_delete(net, grp);
2805 tipc_sk_withdraw(tsk, scope, &seq);
2810 * tipc_setsockopt - set socket option
2811 * @sock: socket structure
2812 * @lvl: option level
2813 * @opt: option identifier
2814 * @ov: pointer to new option value
2815 * @ol: length of option value
2817 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2818 * (to ease compatibility).
2820 * Returns 0 on success, errno otherwise
2822 static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
2823 char __user *ov, unsigned int ol)
2825 struct sock *sk = sock->sk;
2826 struct tipc_sock *tsk = tipc_sk(sk);
2827 struct tipc_group_req mreq;
2831 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2833 if (lvl != SOL_TIPC)
2834 return -ENOPROTOOPT;
2837 case TIPC_IMPORTANCE:
2838 case TIPC_SRC_DROPPABLE:
2839 case TIPC_DEST_DROPPABLE:
2840 case TIPC_CONN_TIMEOUT:
2841 if (ol < sizeof(value))
2843 if (get_user(value, (u32 __user *)ov))
2846 case TIPC_GROUP_JOIN:
2847 if (ol < sizeof(mreq))
2849 if (copy_from_user(&mreq, ov, sizeof(mreq)))
2860 case TIPC_IMPORTANCE:
2861 res = tsk_set_importance(tsk, value);
2863 case TIPC_SRC_DROPPABLE:
2864 if (sock->type != SOCK_STREAM)
2865 tsk_set_unreliable(tsk, value);
2869 case TIPC_DEST_DROPPABLE:
2870 tsk_set_unreturnable(tsk, value);
2872 case TIPC_CONN_TIMEOUT:
2873 tipc_sk(sk)->conn_timeout = value;
2875 case TIPC_MCAST_BROADCAST:
2876 tsk->mc_method.rcast = false;
2877 tsk->mc_method.mandatory = true;
2879 case TIPC_MCAST_REPLICAST:
2880 tsk->mc_method.rcast = true;
2881 tsk->mc_method.mandatory = true;
2883 case TIPC_GROUP_JOIN:
2884 res = tipc_sk_join(tsk, &mreq);
2886 case TIPC_GROUP_LEAVE:
2887 res = tipc_sk_leave(tsk);
2899 * tipc_getsockopt - get socket option
2900 * @sock: socket structure
2901 * @lvl: option level
2902 * @opt: option identifier
2903 * @ov: receptacle for option value
2904 * @ol: receptacle for length of option value
2906 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2907 * (to ease compatibility).
2909 * Returns 0 on success, errno otherwise
2911 static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
2912 char __user *ov, int __user *ol)
2914 struct sock *sk = sock->sk;
2915 struct tipc_sock *tsk = tipc_sk(sk);
2916 struct tipc_name_seq seq;
2921 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
2922 return put_user(0, ol);
2923 if (lvl != SOL_TIPC)
2924 return -ENOPROTOOPT;
2925 res = get_user(len, ol);
2932 case TIPC_IMPORTANCE:
2933 value = tsk_importance(tsk);
2935 case TIPC_SRC_DROPPABLE:
2936 value = tsk_unreliable(tsk);
2938 case TIPC_DEST_DROPPABLE:
2939 value = tsk_unreturnable(tsk);
2941 case TIPC_CONN_TIMEOUT:
2942 value = tsk->conn_timeout;
2943 /* no need to set "res", since already 0 at this point */
2945 case TIPC_NODE_RECVQ_DEPTH:
2946 value = 0; /* was tipc_queue_size, now obsolete */
2948 case TIPC_SOCK_RECVQ_DEPTH:
2949 value = skb_queue_len(&sk->sk_receive_queue);
2951 case TIPC_GROUP_JOIN:
2954 tipc_group_self(tsk->group, &seq, &scope);
2964 return res; /* "get" failed */
2966 if (len < sizeof(value))
2969 if (copy_to_user(ov, &value, sizeof(value)))
2972 return put_user(sizeof(value), ol);
2975 static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2977 struct net *net = sock_net(sock->sk);
2978 struct tipc_sioc_nodeid_req nr = {0};
2979 struct tipc_sioc_ln_req lnr;
2980 void __user *argp = (void __user *)arg;
2983 case SIOCGETLINKNAME:
2984 if (copy_from_user(&lnr, argp, sizeof(lnr)))
2986 if (!tipc_node_get_linkname(net,
2987 lnr.bearer_id & 0xffff, lnr.peer,
2988 lnr.linkname, TIPC_MAX_LINK_NAME)) {
2989 if (copy_to_user(argp, &lnr, sizeof(lnr)))
2993 return -EADDRNOTAVAIL;
2995 if (copy_from_user(&nr, argp, sizeof(nr)))
2997 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
2998 return -EADDRNOTAVAIL;
2999 if (copy_to_user(argp, &nr, sizeof(nr)))
3003 return -ENOIOCTLCMD;
3007 static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3009 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3010 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3011 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3013 tsk1->peer.family = AF_TIPC;
3014 tsk1->peer.addrtype = TIPC_ADDR_ID;
3015 tsk1->peer.scope = TIPC_NODE_SCOPE;
3016 tsk1->peer.addr.id.ref = tsk2->portid;
3017 tsk1->peer.addr.id.node = onode;
3018 tsk2->peer.family = AF_TIPC;
3019 tsk2->peer.addrtype = TIPC_ADDR_ID;
3020 tsk2->peer.scope = TIPC_NODE_SCOPE;
3021 tsk2->peer.addr.id.ref = tsk1->portid;
3022 tsk2->peer.addr.id.node = onode;
3024 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3025 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3029 /* Protocol switches for the various types of TIPC sockets */
3031 static const struct proto_ops msg_ops = {
3032 .owner = THIS_MODULE,
3034 .release = tipc_release,
3036 .connect = tipc_connect,
3037 .socketpair = tipc_socketpair,
3038 .accept = sock_no_accept,
3039 .getname = tipc_getname,
3041 .ioctl = tipc_ioctl,
3042 .listen = sock_no_listen,
3043 .shutdown = tipc_shutdown,
3044 .setsockopt = tipc_setsockopt,
3045 .getsockopt = tipc_getsockopt,
3046 .sendmsg = tipc_sendmsg,
3047 .recvmsg = tipc_recvmsg,
3048 .mmap = sock_no_mmap,
3049 .sendpage = sock_no_sendpage
3052 static const struct proto_ops packet_ops = {
3053 .owner = THIS_MODULE,
3055 .release = tipc_release,
3057 .connect = tipc_connect,
3058 .socketpair = tipc_socketpair,
3059 .accept = tipc_accept,
3060 .getname = tipc_getname,
3062 .ioctl = tipc_ioctl,
3063 .listen = tipc_listen,
3064 .shutdown = tipc_shutdown,
3065 .setsockopt = tipc_setsockopt,
3066 .getsockopt = tipc_getsockopt,
3067 .sendmsg = tipc_send_packet,
3068 .recvmsg = tipc_recvmsg,
3069 .mmap = sock_no_mmap,
3070 .sendpage = sock_no_sendpage
3073 static const struct proto_ops stream_ops = {
3074 .owner = THIS_MODULE,
3076 .release = tipc_release,
3078 .connect = tipc_connect,
3079 .socketpair = tipc_socketpair,
3080 .accept = tipc_accept,
3081 .getname = tipc_getname,
3083 .ioctl = tipc_ioctl,
3084 .listen = tipc_listen,
3085 .shutdown = tipc_shutdown,
3086 .setsockopt = tipc_setsockopt,
3087 .getsockopt = tipc_getsockopt,
3088 .sendmsg = tipc_sendstream,
3089 .recvmsg = tipc_recvstream,
3090 .mmap = sock_no_mmap,
3091 .sendpage = sock_no_sendpage
3094 static const struct net_proto_family tipc_family_ops = {
3095 .owner = THIS_MODULE,
3097 .create = tipc_sk_create
3100 static struct proto tipc_proto = {
3102 .owner = THIS_MODULE,
3103 .obj_size = sizeof(struct tipc_sock),
3104 .sysctl_rmem = sysctl_tipc_rmem
3108 * tipc_socket_init - initialize TIPC socket interface
3110 * Returns 0 on success, errno otherwise
3112 int tipc_socket_init(void)
3116 res = proto_register(&tipc_proto, 1);
3118 pr_err("Failed to register TIPC protocol type\n");
3122 res = sock_register(&tipc_family_ops);
3124 pr_err("Failed to register TIPC socket type\n");
3125 proto_unregister(&tipc_proto);
3133 * tipc_socket_stop - stop TIPC socket interface
3135 void tipc_socket_stop(void)
3137 sock_unregister(tipc_family_ops.family);
3138 proto_unregister(&tipc_proto);
3141 /* Caller should hold socket lock for the passed tipc socket. */
3142 static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3146 struct nlattr *nest;
3148 peer_node = tsk_peer_node(tsk);
3149 peer_port = tsk_peer_port(tsk);
3151 nest = nla_nest_start(skb, TIPC_NLA_SOCK_CON);
3153 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3155 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3158 if (tsk->conn_type != 0) {
3159 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3161 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, tsk->conn_type))
3163 if (nla_put_u32(skb, TIPC_NLA_CON_INST, tsk->conn_instance))
3166 nla_nest_end(skb, nest);
3171 nla_nest_cancel(skb, nest);
3176 static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3179 struct net *net = sock_net(skb->sk);
3180 struct sock *sk = &tsk->sk;
3182 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3183 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3186 if (tipc_sk_connected(sk)) {
3187 if (__tipc_nl_add_sk_con(skb, tsk))
3189 } else if (!list_empty(&tsk->publications)) {
3190 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3196 /* Caller should hold socket lock for the passed tipc socket. */
3197 static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3198 struct tipc_sock *tsk)
3200 struct nlattr *attrs;
3203 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3204 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3208 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3210 goto genlmsg_cancel;
3212 if (__tipc_nl_add_sk_info(skb, tsk))
3213 goto attr_msg_cancel;
3215 nla_nest_end(skb, attrs);
3216 genlmsg_end(skb, hdr);
3221 nla_nest_cancel(skb, attrs);
3223 genlmsg_cancel(skb, hdr);
3228 int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3229 int (*skb_handler)(struct sk_buff *skb,
3230 struct netlink_callback *cb,
3231 struct tipc_sock *tsk))
3233 struct net *net = sock_net(skb->sk);
3234 struct tipc_net *tn = tipc_net(net);
3235 const struct bucket_table *tbl;
3236 u32 prev_portid = cb->args[1];
3237 u32 tbl_id = cb->args[0];
3238 struct rhash_head *pos;
3239 struct tipc_sock *tsk;
3243 tbl = rht_dereference_rcu((&tn->sk_rht)->tbl, &tn->sk_rht);
3244 for (; tbl_id < tbl->size; tbl_id++) {
3245 rht_for_each_entry_rcu(tsk, pos, tbl, tbl_id, node) {
3246 spin_lock_bh(&tsk->sk.sk_lock.slock);
3247 if (prev_portid && prev_portid != tsk->portid) {
3248 spin_unlock_bh(&tsk->sk.sk_lock.slock);
3252 err = skb_handler(skb, cb, tsk);
3254 prev_portid = tsk->portid;
3255 spin_unlock_bh(&tsk->sk.sk_lock.slock);
3260 spin_unlock_bh(&tsk->sk.sk_lock.slock);
3265 cb->args[0] = tbl_id;
3266 cb->args[1] = prev_portid;
3270 EXPORT_SYMBOL(tipc_nl_sk_walk);
3272 int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3273 struct tipc_sock *tsk, u32 sk_filter_state,
3274 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3276 struct sock *sk = &tsk->sk;
3277 struct nlattr *attrs;
3278 struct nlattr *stat;
3280 /*filter response w.r.t sk_state*/
3281 if (!(sk_filter_state & (1 << sk->sk_state)))
3284 attrs = nla_nest_start(skb, TIPC_NLA_SOCK);
3288 if (__tipc_nl_add_sk_info(skb, tsk))
3289 goto attr_msg_cancel;
3291 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3292 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3293 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3294 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3295 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3297 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3298 tipc_diag_gen_cookie(sk),
3300 goto attr_msg_cancel;
3302 stat = nla_nest_start(skb, TIPC_NLA_SOCK_STAT);
3304 goto attr_msg_cancel;
3306 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3307 skb_queue_len(&sk->sk_receive_queue)) ||
3308 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3309 skb_queue_len(&sk->sk_write_queue)) ||
3310 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3311 atomic_read(&sk->sk_drops)))
3312 goto stat_msg_cancel;
3314 if (tsk->cong_link_cnt &&
3315 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3316 goto stat_msg_cancel;
3318 if (tsk_conn_cong(tsk) &&
3319 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3320 goto stat_msg_cancel;
3322 nla_nest_end(skb, stat);
3325 if (tipc_group_fill_sock_diag(tsk->group, skb))
3326 goto stat_msg_cancel;
3328 nla_nest_end(skb, attrs);
3333 nla_nest_cancel(skb, stat);
3335 nla_nest_cancel(skb, attrs);
3339 EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3341 int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3343 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3346 /* Caller should hold socket lock for the passed tipc socket. */
3347 static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3348 struct netlink_callback *cb,
3349 struct publication *publ)
3352 struct nlattr *attrs;
3354 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3355 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3359 attrs = nla_nest_start(skb, TIPC_NLA_PUBL);
3361 goto genlmsg_cancel;
3363 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3364 goto attr_msg_cancel;
3365 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->type))
3366 goto attr_msg_cancel;
3367 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->lower))
3368 goto attr_msg_cancel;
3369 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->upper))
3370 goto attr_msg_cancel;
3372 nla_nest_end(skb, attrs);
3373 genlmsg_end(skb, hdr);
3378 nla_nest_cancel(skb, attrs);
3380 genlmsg_cancel(skb, hdr);
3385 /* Caller should hold socket lock for the passed tipc socket. */
3386 static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3387 struct netlink_callback *cb,
3388 struct tipc_sock *tsk, u32 *last_publ)
3391 struct publication *p;
3394 list_for_each_entry(p, &tsk->publications, binding_sock) {
3395 if (p->key == *last_publ)
3398 if (p->key != *last_publ) {
3399 /* We never set seq or call nl_dump_check_consistent()
3400 * this means that setting prev_seq here will cause the
3401 * consistence check to fail in the netlink callback
3402 * handler. Resulting in the last NLMSG_DONE message
3403 * having the NLM_F_DUMP_INTR flag set.
3410 p = list_first_entry(&tsk->publications, struct publication,
3414 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3415 err = __tipc_nl_add_sk_publ(skb, cb, p);
3417 *last_publ = p->key;
3426 int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3429 u32 tsk_portid = cb->args[0];
3430 u32 last_publ = cb->args[1];
3431 u32 done = cb->args[2];
3432 struct net *net = sock_net(skb->sk);
3433 struct tipc_sock *tsk;
3436 struct nlattr **attrs;
3437 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3439 err = tipc_nlmsg_parse(cb->nlh, &attrs);
3443 if (!attrs[TIPC_NLA_SOCK])
3446 err = nla_parse_nested(sock, TIPC_NLA_SOCK_MAX,
3447 attrs[TIPC_NLA_SOCK],
3448 tipc_nl_sock_policy, NULL);
3452 if (!sock[TIPC_NLA_SOCK_REF])
3455 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3461 tsk = tipc_sk_lookup(net, tsk_portid);
3465 lock_sock(&tsk->sk);
3466 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3469 release_sock(&tsk->sk);
3472 cb->args[0] = tsk_portid;
3473 cb->args[1] = last_publ;