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
)
414 const struct proto_ops
*ops
;
416 struct tipc_sock
*tsk
;
417 struct tipc_msg
*msg
;
419 /* Validate arguments */
420 if (unlikely(protocol
!= 0))
421 return -EPROTONOSUPPORT
;
423 switch (sock
->type
) {
438 /* Allocate socket's protocol area */
439 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
, kern
);
444 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
445 INIT_LIST_HEAD(&tsk
->publications
);
446 INIT_LIST_HEAD(&tsk
->cong_links
);
449 /* Finish initializing socket data structures */
451 sock_init_data(sock
, sk
);
452 tipc_set_sk_state(sk
, TIPC_OPEN
);
453 if (tipc_sk_insert(tsk
)) {
454 pr_warn("Socket create failed; port number exhausted\n");
458 /* Ensure tsk is visible before we read own_addr. */
461 tipc_msg_init(tipc_own_addr(net
), msg
, TIPC_LOW_IMPORTANCE
,
462 TIPC_NAMED_MSG
, NAMED_H_SIZE
, 0);
464 msg_set_origport(msg
, tsk
->portid
);
465 timer_setup(&sk
->sk_timer
, tipc_sk_timeout
, 0);
467 sk
->sk_backlog_rcv
= tipc_sk_backlog_rcv
;
468 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
469 sk
->sk_data_ready
= tipc_data_ready
;
470 sk
->sk_write_space
= tipc_write_space
;
471 sk
->sk_destruct
= tipc_sock_destruct
;
472 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
473 tsk
->group_is_open
= true;
474 atomic_set(&tsk
->dupl_rcvcnt
, 0);
476 /* Start out with safe limits until we receive an advertised window */
477 tsk
->snd_win
= tsk_adv_blocks(RCVBUF_MIN
);
478 tsk
->rcv_win
= tsk
->snd_win
;
480 if (tipc_sk_type_connectionless(sk
)) {
481 tsk_set_unreturnable(tsk
, true);
482 if (sock
->type
== SOCK_DGRAM
)
483 tsk_set_unreliable(tsk
, true);
489 static void tipc_sk_callback(struct rcu_head
*head
)
491 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
496 /* Caller should hold socket lock for the socket. */
497 static void __tipc_shutdown(struct socket
*sock
, int error
)
499 struct sock
*sk
= sock
->sk
;
500 struct tipc_sock
*tsk
= tipc_sk(sk
);
501 struct net
*net
= sock_net(sk
);
502 long timeout
= CONN_TIMEOUT_DEFAULT
;
503 u32 dnode
= tsk_peer_node(tsk
);
506 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
507 tipc_wait_for_cond(sock
, &timeout
, (!tsk
->cong_link_cnt
&&
508 !tsk_conn_cong(tsk
)));
510 /* Reject all unreceived messages, except on an active connection
511 * (which disconnects locally & sends a 'FIN+' to peer).
513 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)) != NULL
) {
514 if (TIPC_SKB_CB(skb
)->bytes_read
) {
518 if (!tipc_sk_type_connectionless(sk
) &&
519 sk
->sk_state
!= TIPC_DISCONNECTING
) {
520 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
521 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
523 tipc_sk_respond(sk
, skb
, error
);
526 if (tipc_sk_type_connectionless(sk
))
529 if (sk
->sk_state
!= TIPC_DISCONNECTING
) {
530 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
531 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
532 tsk_own_node(tsk
), tsk_peer_port(tsk
),
535 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
536 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
537 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
542 * tipc_release - destroy a TIPC socket
543 * @sock: socket to destroy
545 * This routine cleans up any messages that are still queued on the socket.
546 * For DGRAM and RDM socket types, all queued messages are rejected.
547 * For SEQPACKET and STREAM socket types, the first message is rejected
548 * and any others are discarded. (If the first message on a STREAM socket
549 * is partially-read, it is discarded and the next one is rejected instead.)
551 * NOTE: Rejected messages are not necessarily returned to the sender! They
552 * are returned or discarded according to the "destination droppable" setting
553 * specified for the message by the sender.
555 * Returns 0 on success, errno otherwise
557 static int tipc_release(struct socket
*sock
)
559 struct sock
*sk
= sock
->sk
;
560 struct tipc_sock
*tsk
;
563 * Exit if socket isn't fully initialized (occurs when a failed accept()
564 * releases a pre-allocated child socket that was never used)
572 __tipc_shutdown(sock
, TIPC_ERR_NO_PORT
);
573 sk
->sk_shutdown
= SHUTDOWN_MASK
;
575 tipc_sk_withdraw(tsk
, 0, NULL
);
576 sk_stop_timer(sk
, &sk
->sk_timer
);
579 /* Reject any messages that accumulated in backlog queue */
581 tipc_dest_list_purge(&tsk
->cong_links
);
582 tsk
->cong_link_cnt
= 0;
583 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
590 * tipc_bind - associate or disassocate TIPC name(s) with a socket
591 * @sock: socket structure
592 * @uaddr: socket address describing name(s) and desired operation
593 * @uaddr_len: size of socket address data structure
595 * Name and name sequence binding is indicated using a positive scope value;
596 * a negative scope value unbinds the specified name. Specifying no name
597 * (i.e. a socket address length of 0) unbinds all names from the socket.
599 * Returns 0 on success, errno otherwise
601 * NOTE: This routine doesn't need to take the socket lock since it doesn't
602 * access any non-constant socket information.
604 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
607 struct sock
*sk
= sock
->sk
;
608 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
609 struct tipc_sock
*tsk
= tipc_sk(sk
);
613 if (unlikely(!uaddr_len
)) {
614 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
621 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
625 if (addr
->family
!= AF_TIPC
) {
630 if (addr
->addrtype
== TIPC_ADDR_NAME
)
631 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
632 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
637 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
638 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
639 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
644 res
= (addr
->scope
>= 0) ?
645 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
646 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
653 * tipc_getname - get port ID of socket or peer socket
654 * @sock: socket structure
655 * @uaddr: area for returned socket address
656 * @uaddr_len: area for returned length of socket address
657 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
659 * Returns 0 on success, errno otherwise
661 * NOTE: This routine doesn't need to take the socket lock since it only
662 * accesses socket information that is unchanging (or which changes in
663 * a completely predictable manner).
665 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
668 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
669 struct sock
*sk
= sock
->sk
;
670 struct tipc_sock
*tsk
= tipc_sk(sk
);
672 memset(addr
, 0, sizeof(*addr
));
674 if ((!tipc_sk_connected(sk
)) &&
675 ((peer
!= 2) || (sk
->sk_state
!= TIPC_DISCONNECTING
)))
677 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
678 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
680 addr
->addr
.id
.ref
= tsk
->portid
;
681 addr
->addr
.id
.node
= tipc_own_addr(sock_net(sk
));
684 addr
->addrtype
= TIPC_ADDR_ID
;
685 addr
->family
= AF_TIPC
;
687 addr
->addr
.name
.domain
= 0;
689 return sizeof(*addr
);
693 * tipc_poll - read and possibly block on pollmask
694 * @file: file structure associated with the socket
695 * @sock: socket for which to calculate the poll bits
698 * Returns pollmask value
701 * It appears that the usual socket locking mechanisms are not useful here
702 * since the pollmask info is potentially out-of-date the moment this routine
703 * exits. TCP and other protocols seem to rely on higher level poll routines
704 * to handle any preventable race conditions, so TIPC will do the same ...
706 * IMPORTANT: The fact that a read or write operation is indicated does NOT
707 * imply that the operation will succeed, merely that it should be performed
708 * and will not block.
710 static __poll_t
tipc_poll(struct file
*file
, struct socket
*sock
,
713 struct sock
*sk
= sock
->sk
;
714 struct tipc_sock
*tsk
= tipc_sk(sk
);
715 __poll_t revents
= 0;
717 sock_poll_wait(file
, wait
);
719 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
720 revents
|= EPOLLRDHUP
| EPOLLIN
| EPOLLRDNORM
;
721 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
724 switch (sk
->sk_state
) {
725 case TIPC_ESTABLISHED
:
726 case TIPC_CONNECTING
:
727 if (!tsk
->cong_link_cnt
&& !tsk_conn_cong(tsk
))
731 if (!skb_queue_empty(&sk
->sk_receive_queue
))
732 revents
|= EPOLLIN
| EPOLLRDNORM
;
735 if (tsk
->group_is_open
&& !tsk
->cong_link_cnt
)
737 if (!tipc_sk_type_connectionless(sk
))
739 if (skb_queue_empty(&sk
->sk_receive_queue
))
741 revents
|= EPOLLIN
| EPOLLRDNORM
;
743 case TIPC_DISCONNECTING
:
744 revents
= EPOLLIN
| EPOLLRDNORM
| EPOLLHUP
;
751 * tipc_sendmcast - send multicast message
752 * @sock: socket structure
753 * @seq: destination address
754 * @msg: message to send
755 * @dlen: length of data to send
756 * @timeout: timeout to wait for wakeup
758 * Called from function tipc_sendmsg(), which has done all sanity checks
759 * Returns the number of bytes sent on success, or errno
761 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
762 struct msghdr
*msg
, size_t dlen
, long timeout
)
764 struct sock
*sk
= sock
->sk
;
765 struct tipc_sock
*tsk
= tipc_sk(sk
);
766 struct tipc_msg
*hdr
= &tsk
->phdr
;
767 struct net
*net
= sock_net(sk
);
768 int mtu
= tipc_bcast_get_mtu(net
);
769 struct tipc_mc_method
*method
= &tsk
->mc_method
;
770 struct sk_buff_head pkts
;
771 struct tipc_nlist dsts
;
777 /* Block or return if any destination link is congested */
778 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
);
782 /* Lookup destination nodes */
783 tipc_nlist_init(&dsts
, tipc_own_addr(net
));
784 tipc_nametbl_lookup_dst_nodes(net
, seq
->type
, seq
->lower
,
786 if (!dsts
.local
&& !dsts
.remote
)
787 return -EHOSTUNREACH
;
789 /* Build message header */
790 msg_set_type(hdr
, TIPC_MCAST_MSG
);
791 msg_set_hdr_sz(hdr
, MCAST_H_SIZE
);
792 msg_set_lookup_scope(hdr
, TIPC_CLUSTER_SCOPE
);
793 msg_set_destport(hdr
, 0);
794 msg_set_destnode(hdr
, 0);
795 msg_set_nametype(hdr
, seq
->type
);
796 msg_set_namelower(hdr
, seq
->lower
);
797 msg_set_nameupper(hdr
, seq
->upper
);
799 /* Build message as chain of buffers */
800 skb_queue_head_init(&pkts
);
801 rc
= tipc_msg_build(hdr
, msg
, 0, dlen
, mtu
, &pkts
);
803 /* Send message if build was successful */
804 if (unlikely(rc
== dlen
))
805 rc
= tipc_mcast_xmit(net
, &pkts
, method
, &dsts
,
806 &tsk
->cong_link_cnt
);
808 tipc_nlist_purge(&dsts
);
810 return rc
? rc
: dlen
;
814 * tipc_send_group_msg - send a message to a member in the group
815 * @net: network namespace
816 * @m: message to send
818 * @dnode: destination node
819 * @dport: destination port
820 * @dlen: total length of message data
822 static int tipc_send_group_msg(struct net
*net
, struct tipc_sock
*tsk
,
823 struct msghdr
*m
, struct tipc_member
*mb
,
824 u32 dnode
, u32 dport
, int dlen
)
826 u16 bc_snd_nxt
= tipc_group_bc_snd_nxt(tsk
->group
);
827 struct tipc_mc_method
*method
= &tsk
->mc_method
;
828 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
829 struct tipc_msg
*hdr
= &tsk
->phdr
;
830 struct sk_buff_head pkts
;
833 /* Complete message header */
834 msg_set_type(hdr
, TIPC_GRP_UCAST_MSG
);
835 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
836 msg_set_destport(hdr
, dport
);
837 msg_set_destnode(hdr
, dnode
);
838 msg_set_grp_bc_seqno(hdr
, bc_snd_nxt
);
840 /* Build message as chain of buffers */
841 skb_queue_head_init(&pkts
);
842 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
843 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
844 if (unlikely(rc
!= dlen
))
848 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
849 if (unlikely(rc
== -ELINKCONG
)) {
850 tipc_dest_push(&tsk
->cong_links
, dnode
, 0);
851 tsk
->cong_link_cnt
++;
854 /* Update send window */
855 tipc_group_update_member(mb
, blks
);
857 /* A broadcast sent within next EXPIRE period must follow same path */
858 method
->rcast
= true;
859 method
->mandatory
= true;
864 * tipc_send_group_unicast - send message to a member in the group
865 * @sock: socket structure
866 * @m: message to send
867 * @dlen: total length of message data
868 * @timeout: timeout to wait for wakeup
870 * Called from function tipc_sendmsg(), which has done all sanity checks
871 * Returns the number of bytes sent on success, or errno
873 static int tipc_send_group_unicast(struct socket
*sock
, struct msghdr
*m
,
874 int dlen
, long timeout
)
876 struct sock
*sk
= sock
->sk
;
877 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
878 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
879 struct tipc_sock
*tsk
= tipc_sk(sk
);
880 struct tipc_group
*grp
= tsk
->group
;
881 struct net
*net
= sock_net(sk
);
882 struct tipc_member
*mb
= NULL
;
886 node
= dest
->addr
.id
.node
;
887 port
= dest
->addr
.id
.ref
;
889 return -EHOSTUNREACH
;
891 /* Block or return if destination link or member is congested */
892 rc
= tipc_wait_for_cond(sock
, &timeout
,
893 !tipc_dest_find(&tsk
->cong_links
, node
, 0) &&
894 !tipc_group_cong(grp
, node
, port
, blks
, &mb
));
899 return -EHOSTUNREACH
;
901 rc
= tipc_send_group_msg(net
, tsk
, m
, mb
, node
, port
, dlen
);
903 return rc
? rc
: dlen
;
907 * tipc_send_group_anycast - send message to any member with given identity
908 * @sock: socket structure
909 * @m: message to send
910 * @dlen: total length of message data
911 * @timeout: timeout to wait for wakeup
913 * Called from function tipc_sendmsg(), which has done all sanity checks
914 * Returns the number of bytes sent on success, or errno
916 static int tipc_send_group_anycast(struct socket
*sock
, struct msghdr
*m
,
917 int dlen
, long timeout
)
919 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
920 struct sock
*sk
= sock
->sk
;
921 struct tipc_sock
*tsk
= tipc_sk(sk
);
922 struct list_head
*cong_links
= &tsk
->cong_links
;
923 int blks
= tsk_blocks(GROUP_H_SIZE
+ dlen
);
924 struct tipc_group
*grp
= tsk
->group
;
925 struct tipc_msg
*hdr
= &tsk
->phdr
;
926 struct tipc_member
*first
= NULL
;
927 struct tipc_member
*mbr
= NULL
;
928 struct net
*net
= sock_net(sk
);
929 u32 node
, port
, exclude
;
930 struct list_head dsts
;
931 u32 type
, inst
, scope
;
936 INIT_LIST_HEAD(&dsts
);
938 type
= msg_nametype(hdr
);
939 inst
= dest
->addr
.name
.name
.instance
;
940 scope
= msg_lookup_scope(hdr
);
941 exclude
= tipc_group_exclude(grp
);
943 while (++lookups
< 4) {
946 /* Look for a non-congested destination member, if any */
948 if (!tipc_nametbl_lookup(net
, type
, inst
, scope
, &dsts
,
949 &dstcnt
, exclude
, false))
950 return -EHOSTUNREACH
;
951 tipc_dest_pop(&dsts
, &node
, &port
);
952 cong
= tipc_group_cong(grp
, node
, port
, blks
, &mbr
);
961 /* Start over if destination was not in member list */
965 if (likely(!cong
&& !tipc_dest_find(cong_links
, node
, 0)))
968 /* Block or return if destination link or member is congested */
969 rc
= tipc_wait_for_cond(sock
, &timeout
,
970 !tipc_dest_find(cong_links
, node
, 0) &&
971 !tipc_group_cong(grp
, node
, port
,
976 /* Send, unless destination disappeared while waiting */
981 if (unlikely(lookups
>= 4))
982 return -EHOSTUNREACH
;
984 rc
= tipc_send_group_msg(net
, tsk
, m
, mbr
, node
, port
, dlen
);
986 return rc
? rc
: dlen
;
990 * tipc_send_group_bcast - send message to all members in communication group
991 * @sk: socket structure
992 * @m: message to send
993 * @dlen: total length of message data
994 * @timeout: timeout to wait for wakeup
996 * Called from function tipc_sendmsg(), which has done all sanity checks
997 * Returns the number of bytes sent on success, or errno
999 static int tipc_send_group_bcast(struct socket
*sock
, struct msghdr
*m
,
1000 int dlen
, long timeout
)
1002 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1003 struct sock
*sk
= sock
->sk
;
1004 struct net
*net
= sock_net(sk
);
1005 struct tipc_sock
*tsk
= tipc_sk(sk
);
1006 struct tipc_group
*grp
= tsk
->group
;
1007 struct tipc_nlist
*dsts
= tipc_group_dests(grp
);
1008 struct tipc_mc_method
*method
= &tsk
->mc_method
;
1009 bool ack
= method
->mandatory
&& method
->rcast
;
1010 int blks
= tsk_blocks(MCAST_H_SIZE
+ dlen
);
1011 struct tipc_msg
*hdr
= &tsk
->phdr
;
1012 int mtu
= tipc_bcast_get_mtu(net
);
1013 struct sk_buff_head pkts
;
1014 int rc
= -EHOSTUNREACH
;
1016 if (!dsts
->local
&& !dsts
->remote
)
1017 return -EHOSTUNREACH
;
1019 /* Block or return if any destination link or member is congested */
1020 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
&&
1021 !tipc_group_bc_cong(grp
, blks
));
1025 /* Complete message header */
1027 msg_set_type(hdr
, TIPC_GRP_MCAST_MSG
);
1028 msg_set_nameinst(hdr
, dest
->addr
.name
.name
.instance
);
1030 msg_set_type(hdr
, TIPC_GRP_BCAST_MSG
);
1031 msg_set_nameinst(hdr
, 0);
1033 msg_set_hdr_sz(hdr
, GROUP_H_SIZE
);
1034 msg_set_destport(hdr
, 0);
1035 msg_set_destnode(hdr
, 0);
1036 msg_set_grp_bc_seqno(hdr
, tipc_group_bc_snd_nxt(grp
));
1038 /* Avoid getting stuck with repeated forced replicasts */
1039 msg_set_grp_bc_ack_req(hdr
, ack
);
1041 /* Build message as chain of buffers */
1042 skb_queue_head_init(&pkts
);
1043 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1044 if (unlikely(rc
!= dlen
))
1048 rc
= tipc_mcast_xmit(net
, &pkts
, method
, dsts
, &tsk
->cong_link_cnt
);
1052 /* Update broadcast sequence number and send windows */
1053 tipc_group_update_bc_members(tsk
->group
, blks
, ack
);
1055 /* Broadcast link is now free to choose method for next broadcast */
1056 method
->mandatory
= false;
1057 method
->expires
= jiffies
;
1063 * tipc_send_group_mcast - send message to all members with given identity
1064 * @sock: socket structure
1065 * @m: message to send
1066 * @dlen: total length of message data
1067 * @timeout: timeout to wait for wakeup
1069 * Called from function tipc_sendmsg(), which has done all sanity checks
1070 * Returns the number of bytes sent on success, or errno
1072 static int tipc_send_group_mcast(struct socket
*sock
, struct msghdr
*m
,
1073 int dlen
, long timeout
)
1075 struct sock
*sk
= sock
->sk
;
1076 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1077 struct tipc_sock
*tsk
= tipc_sk(sk
);
1078 struct tipc_group
*grp
= tsk
->group
;
1079 struct tipc_msg
*hdr
= &tsk
->phdr
;
1080 struct net
*net
= sock_net(sk
);
1081 u32 type
, inst
, scope
, exclude
;
1082 struct list_head dsts
;
1085 INIT_LIST_HEAD(&dsts
);
1087 type
= msg_nametype(hdr
);
1088 inst
= dest
->addr
.name
.name
.instance
;
1089 scope
= msg_lookup_scope(hdr
);
1090 exclude
= tipc_group_exclude(grp
);
1092 if (!tipc_nametbl_lookup(net
, type
, inst
, scope
, &dsts
,
1093 &dstcnt
, exclude
, true))
1094 return -EHOSTUNREACH
;
1097 tipc_dest_pop(&dsts
, &dest
->addr
.id
.node
, &dest
->addr
.id
.ref
);
1098 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1101 tipc_dest_list_purge(&dsts
);
1102 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1106 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1107 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1108 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1110 * Multi-threaded: parallel calls with reference to same queues may occur
1112 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
1113 struct sk_buff_head
*inputq
)
1115 u32 self
= tipc_own_addr(net
);
1116 u32 type
, lower
, upper
, scope
;
1117 struct sk_buff
*skb
, *_skb
;
1119 struct sk_buff_head tmpq
;
1120 struct list_head dports
;
1121 struct tipc_msg
*hdr
;
1122 int user
, mtyp
, hlen
;
1125 __skb_queue_head_init(&tmpq
);
1126 INIT_LIST_HEAD(&dports
);
1128 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
1129 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
1131 user
= msg_user(hdr
);
1132 mtyp
= msg_type(hdr
);
1133 hlen
= skb_headroom(skb
) + msg_hdr_sz(hdr
);
1134 onode
= msg_orignode(hdr
);
1135 type
= msg_nametype(hdr
);
1137 if (mtyp
== TIPC_GRP_UCAST_MSG
|| user
== GROUP_PROTOCOL
) {
1138 spin_lock_bh(&inputq
->lock
);
1139 if (skb_peek(arrvq
) == skb
) {
1140 __skb_dequeue(arrvq
);
1141 __skb_queue_tail(inputq
, skb
);
1144 spin_unlock_bh(&inputq
->lock
);
1148 /* Group messages require exact scope match */
1149 if (msg_in_group(hdr
)) {
1152 scope
= msg_lookup_scope(hdr
);
1155 /* TIPC_NODE_SCOPE means "any scope" in this context */
1157 scope
= TIPC_NODE_SCOPE
;
1159 scope
= TIPC_CLUSTER_SCOPE
;
1161 lower
= msg_namelower(hdr
);
1162 upper
= msg_nameupper(hdr
);
1165 /* Create destination port list: */
1166 tipc_nametbl_mc_lookup(net
, type
, lower
, upper
,
1167 scope
, exact
, &dports
);
1169 /* Clone message per destination */
1170 while (tipc_dest_pop(&dports
, NULL
, &portid
)) {
1171 _skb
= __pskb_copy(skb
, hlen
, GFP_ATOMIC
);
1173 msg_set_destport(buf_msg(_skb
), portid
);
1174 __skb_queue_tail(&tmpq
, _skb
);
1177 pr_warn("Failed to clone mcast rcv buffer\n");
1179 /* Append to inputq if not already done by other thread */
1180 spin_lock_bh(&inputq
->lock
);
1181 if (skb_peek(arrvq
) == skb
) {
1182 skb_queue_splice_tail_init(&tmpq
, inputq
);
1183 kfree_skb(__skb_dequeue(arrvq
));
1185 spin_unlock_bh(&inputq
->lock
);
1186 __skb_queue_purge(&tmpq
);
1189 tipc_sk_rcv(net
, inputq
);
1193 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1194 * @tsk: receiving socket
1195 * @skb: pointer to message buffer.
1197 static void tipc_sk_conn_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
,
1198 struct sk_buff_head
*xmitq
)
1200 struct tipc_msg
*hdr
= buf_msg(skb
);
1201 u32 onode
= tsk_own_node(tsk
);
1202 struct sock
*sk
= &tsk
->sk
;
1203 int mtyp
= msg_type(hdr
);
1206 /* Ignore if connection cannot be validated: */
1207 if (!tsk_peer_msg(tsk
, hdr
))
1210 if (unlikely(msg_errcode(hdr
))) {
1211 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1212 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
1213 tsk_peer_port(tsk
));
1214 sk
->sk_state_change(sk
);
1218 tsk
->probe_unacked
= false;
1220 if (mtyp
== CONN_PROBE
) {
1221 msg_set_type(hdr
, CONN_PROBE_REPLY
);
1222 if (tipc_msg_reverse(onode
, &skb
, TIPC_OK
))
1223 __skb_queue_tail(xmitq
, skb
);
1225 } else if (mtyp
== CONN_ACK
) {
1226 conn_cong
= tsk_conn_cong(tsk
);
1227 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
1228 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1229 tsk
->snd_win
= msg_adv_win(hdr
);
1231 sk
->sk_write_space(sk
);
1232 } else if (mtyp
!= CONN_PROBE_REPLY
) {
1233 pr_warn("Received unknown CONN_PROTO msg\n");
1240 * tipc_sendmsg - send message in connectionless manner
1241 * @sock: socket structure
1242 * @m: message to send
1243 * @dsz: amount of user data to be sent
1245 * Message must have an destination specified explicitly.
1246 * Used for SOCK_RDM and SOCK_DGRAM messages,
1247 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1248 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1250 * Returns the number of bytes sent on success, or errno otherwise
1252 static int tipc_sendmsg(struct socket
*sock
,
1253 struct msghdr
*m
, size_t dsz
)
1255 struct sock
*sk
= sock
->sk
;
1259 ret
= __tipc_sendmsg(sock
, m
, dsz
);
1265 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1267 struct sock
*sk
= sock
->sk
;
1268 struct net
*net
= sock_net(sk
);
1269 struct tipc_sock
*tsk
= tipc_sk(sk
);
1270 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1271 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1272 struct list_head
*clinks
= &tsk
->cong_links
;
1273 bool syn
= !tipc_sk_type_connectionless(sk
);
1274 struct tipc_group
*grp
= tsk
->group
;
1275 struct tipc_msg
*hdr
= &tsk
->phdr
;
1276 struct tipc_name_seq
*seq
;
1277 struct sk_buff_head pkts
;
1278 u32 dport
, dnode
= 0;
1282 if (unlikely(dlen
> TIPC_MAX_USER_MSG_SIZE
))
1286 if (unlikely(m
->msg_namelen
< sizeof(*dest
)))
1288 if (unlikely(dest
->family
!= AF_TIPC
))
1294 return tipc_send_group_bcast(sock
, m
, dlen
, timeout
);
1295 if (dest
->addrtype
== TIPC_ADDR_NAME
)
1296 return tipc_send_group_anycast(sock
, m
, dlen
, timeout
);
1297 if (dest
->addrtype
== TIPC_ADDR_ID
)
1298 return tipc_send_group_unicast(sock
, m
, dlen
, timeout
);
1299 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1300 return tipc_send_group_mcast(sock
, m
, dlen
, timeout
);
1304 if (unlikely(!dest
)) {
1306 if (!syn
|| dest
->family
!= AF_TIPC
)
1307 return -EDESTADDRREQ
;
1310 if (unlikely(syn
)) {
1311 if (sk
->sk_state
== TIPC_LISTEN
)
1313 if (sk
->sk_state
!= TIPC_OPEN
)
1317 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1318 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
1319 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
1323 seq
= &dest
->addr
.nameseq
;
1324 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
1325 return tipc_sendmcast(sock
, seq
, m
, dlen
, timeout
);
1327 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
1328 type
= dest
->addr
.name
.name
.type
;
1329 inst
= dest
->addr
.name
.name
.instance
;
1330 dnode
= dest
->addr
.name
.domain
;
1331 msg_set_type(hdr
, TIPC_NAMED_MSG
);
1332 msg_set_hdr_sz(hdr
, NAMED_H_SIZE
);
1333 msg_set_nametype(hdr
, type
);
1334 msg_set_nameinst(hdr
, inst
);
1335 msg_set_lookup_scope(hdr
, tipc_node2scope(dnode
));
1336 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
1337 msg_set_destnode(hdr
, dnode
);
1338 msg_set_destport(hdr
, dport
);
1339 if (unlikely(!dport
&& !dnode
))
1340 return -EHOSTUNREACH
;
1341 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
1342 dnode
= dest
->addr
.id
.node
;
1343 msg_set_type(hdr
, TIPC_DIRECT_MSG
);
1344 msg_set_lookup_scope(hdr
, 0);
1345 msg_set_destnode(hdr
, dnode
);
1346 msg_set_destport(hdr
, dest
->addr
.id
.ref
);
1347 msg_set_hdr_sz(hdr
, BASIC_H_SIZE
);
1352 /* Block or return if destination link is congested */
1353 rc
= tipc_wait_for_cond(sock
, &timeout
,
1354 !tipc_dest_find(clinks
, dnode
, 0));
1358 skb_queue_head_init(&pkts
);
1359 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
1360 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
1361 if (unlikely(rc
!= dlen
))
1364 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1365 if (unlikely(rc
== -ELINKCONG
)) {
1366 tipc_dest_push(clinks
, dnode
, 0);
1367 tsk
->cong_link_cnt
++;
1371 if (unlikely(syn
&& !rc
))
1372 tipc_set_sk_state(sk
, TIPC_CONNECTING
);
1374 return rc
? rc
: dlen
;
1378 * tipc_sendstream - send stream-oriented data
1379 * @sock: socket structure
1381 * @dsz: total length of data to be transmitted
1383 * Used for SOCK_STREAM data.
1385 * Returns the number of bytes sent on success (or partial success),
1386 * or errno if no data sent
1388 static int tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1390 struct sock
*sk
= sock
->sk
;
1394 ret
= __tipc_sendstream(sock
, m
, dsz
);
1400 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1402 struct sock
*sk
= sock
->sk
;
1403 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1404 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1405 struct tipc_sock
*tsk
= tipc_sk(sk
);
1406 struct tipc_msg
*hdr
= &tsk
->phdr
;
1407 struct net
*net
= sock_net(sk
);
1408 struct sk_buff_head pkts
;
1409 u32 dnode
= tsk_peer_node(tsk
);
1413 skb_queue_head_init(&pkts
);
1415 if (unlikely(dlen
> INT_MAX
))
1418 /* Handle implicit connection setup */
1419 if (unlikely(dest
)) {
1420 rc
= __tipc_sendmsg(sock
, m
, dlen
);
1421 if (dlen
&& (dlen
== rc
))
1422 tsk
->snt_unacked
= tsk_inc(tsk
, dlen
+ msg_hdr_sz(hdr
));
1427 rc
= tipc_wait_for_cond(sock
, &timeout
,
1428 (!tsk
->cong_link_cnt
&&
1429 !tsk_conn_cong(tsk
) &&
1430 tipc_sk_connected(sk
)));
1434 send
= min_t(size_t, dlen
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1435 rc
= tipc_msg_build(hdr
, m
, sent
, send
, tsk
->max_pkt
, &pkts
);
1436 if (unlikely(rc
!= send
))
1439 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1440 if (unlikely(rc
== -ELINKCONG
)) {
1441 tsk
->cong_link_cnt
= 1;
1445 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ MIN_H_SIZE
);
1448 } while (sent
< dlen
&& !rc
);
1450 return sent
? sent
: rc
;
1454 * tipc_send_packet - send a connection-oriented message
1455 * @sock: socket structure
1456 * @m: message to send
1457 * @dsz: length of data to be transmitted
1459 * Used for SOCK_SEQPACKET messages.
1461 * Returns the number of bytes sent on success, or errno otherwise
1463 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1465 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1468 return tipc_sendstream(sock
, m
, dsz
);
1471 /* tipc_sk_finish_conn - complete the setup of a connection
1473 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1476 struct sock
*sk
= &tsk
->sk
;
1477 struct net
*net
= sock_net(sk
);
1478 struct tipc_msg
*msg
= &tsk
->phdr
;
1480 msg_set_destnode(msg
, peer_node
);
1481 msg_set_destport(msg
, peer_port
);
1482 msg_set_type(msg
, TIPC_CONN_MSG
);
1483 msg_set_lookup_scope(msg
, 0);
1484 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1486 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
1487 tipc_set_sk_state(sk
, TIPC_ESTABLISHED
);
1488 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1489 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1490 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1491 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1494 /* Fall back to message based flow control */
1495 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1496 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1500 * tipc_sk_set_orig_addr - capture sender's address for received message
1501 * @m: descriptor for message info
1502 * @hdr: received message header
1504 * Note: Address is not captured if not requested by receiver.
1506 static void tipc_sk_set_orig_addr(struct msghdr
*m
, struct sk_buff
*skb
)
1508 DECLARE_SOCKADDR(struct sockaddr_pair
*, srcaddr
, m
->msg_name
);
1509 struct tipc_msg
*hdr
= buf_msg(skb
);
1514 srcaddr
->sock
.family
= AF_TIPC
;
1515 srcaddr
->sock
.addrtype
= TIPC_ADDR_ID
;
1516 srcaddr
->sock
.scope
= 0;
1517 srcaddr
->sock
.addr
.id
.ref
= msg_origport(hdr
);
1518 srcaddr
->sock
.addr
.id
.node
= msg_orignode(hdr
);
1519 srcaddr
->sock
.addr
.name
.domain
= 0;
1520 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1522 if (!msg_in_group(hdr
))
1525 /* Group message users may also want to know sending member's id */
1526 srcaddr
->member
.family
= AF_TIPC
;
1527 srcaddr
->member
.addrtype
= TIPC_ADDR_NAME
;
1528 srcaddr
->member
.scope
= 0;
1529 srcaddr
->member
.addr
.name
.name
.type
= msg_nametype(hdr
);
1530 srcaddr
->member
.addr
.name
.name
.instance
= TIPC_SKB_CB(skb
)->orig_member
;
1531 srcaddr
->member
.addr
.name
.domain
= 0;
1532 m
->msg_namelen
= sizeof(*srcaddr
);
1536 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1537 * @m: descriptor for message info
1538 * @msg: received message header
1539 * @tsk: TIPC port associated with message
1541 * Note: Ancillary data is not captured if not requested by receiver.
1543 * Returns 0 if successful, otherwise errno
1545 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1546 struct tipc_sock
*tsk
)
1554 if (likely(m
->msg_controllen
== 0))
1557 /* Optionally capture errored message object(s) */
1558 err
= msg
? msg_errcode(msg
) : 0;
1559 if (unlikely(err
)) {
1561 anc_data
[1] = msg_data_sz(msg
);
1562 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1566 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1573 /* Optionally capture message destination object */
1574 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1575 switch (dest_type
) {
1576 case TIPC_NAMED_MSG
:
1578 anc_data
[0] = msg_nametype(msg
);
1579 anc_data
[1] = msg_namelower(msg
);
1580 anc_data
[2] = msg_namelower(msg
);
1582 case TIPC_MCAST_MSG
:
1584 anc_data
[0] = msg_nametype(msg
);
1585 anc_data
[1] = msg_namelower(msg
);
1586 anc_data
[2] = msg_nameupper(msg
);
1589 has_name
= (tsk
->conn_type
!= 0);
1590 anc_data
[0] = tsk
->conn_type
;
1591 anc_data
[1] = tsk
->conn_instance
;
1592 anc_data
[2] = tsk
->conn_instance
;
1598 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1606 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1608 struct sock
*sk
= &tsk
->sk
;
1609 struct net
*net
= sock_net(sk
);
1610 struct sk_buff
*skb
= NULL
;
1611 struct tipc_msg
*msg
;
1612 u32 peer_port
= tsk_peer_port(tsk
);
1613 u32 dnode
= tsk_peer_node(tsk
);
1615 if (!tipc_sk_connected(sk
))
1617 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1618 dnode
, tsk_own_node(tsk
), peer_port
,
1619 tsk
->portid
, TIPC_OK
);
1623 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1624 tsk
->rcv_unacked
= 0;
1626 /* Adjust to and advertize the correct window limit */
1627 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1628 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1629 msg_set_adv_win(msg
, tsk
->rcv_win
);
1631 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1634 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1636 struct sock
*sk
= sock
->sk
;
1638 long timeo
= *timeop
;
1639 int err
= sock_error(sk
);
1645 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1646 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1647 if (sk
->sk_shutdown
& RCV_SHUTDOWN
) {
1652 timeo
= schedule_timeout(timeo
);
1656 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1661 err
= sock_intr_errno(timeo
);
1662 if (signal_pending(current
))
1665 err
= sock_error(sk
);
1669 finish_wait(sk_sleep(sk
), &wait
);
1675 * tipc_recvmsg - receive packet-oriented message
1676 * @m: descriptor for message info
1677 * @buflen: length of user buffer area
1678 * @flags: receive flags
1680 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1681 * If the complete message doesn't fit in user area, truncate it.
1683 * Returns size of returned message data, errno otherwise
1685 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
,
1686 size_t buflen
, int flags
)
1688 struct sock
*sk
= sock
->sk
;
1689 bool connected
= !tipc_sk_type_connectionless(sk
);
1690 struct tipc_sock
*tsk
= tipc_sk(sk
);
1691 int rc
, err
, hlen
, dlen
, copy
;
1692 struct sk_buff_head xmitq
;
1693 struct tipc_msg
*hdr
;
1694 struct sk_buff
*skb
;
1698 /* Catch invalid receive requests */
1699 if (unlikely(!buflen
))
1703 if (unlikely(connected
&& sk
->sk_state
== TIPC_OPEN
)) {
1707 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1709 /* Step rcv queue to first msg with data or error; wait if necessary */
1711 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1714 skb
= skb_peek(&sk
->sk_receive_queue
);
1716 dlen
= msg_data_sz(hdr
);
1717 hlen
= msg_hdr_sz(hdr
);
1718 err
= msg_errcode(hdr
);
1719 grp_evt
= msg_is_grp_evt(hdr
);
1720 if (likely(dlen
|| err
))
1722 tsk_advance_rx_queue(sk
);
1725 /* Collect msg meta data, including error code and rejected data */
1726 tipc_sk_set_orig_addr(m
, skb
);
1727 rc
= tipc_sk_anc_data_recv(m
, hdr
, tsk
);
1731 /* Capture data if non-error msg, otherwise just set return value */
1733 copy
= min_t(int, dlen
, buflen
);
1734 if (unlikely(copy
!= dlen
))
1735 m
->msg_flags
|= MSG_TRUNC
;
1736 rc
= skb_copy_datagram_msg(skb
, hlen
, m
, copy
);
1740 if (err
!= TIPC_CONN_SHUTDOWN
&& connected
&& !m
->msg_control
)
1746 /* Mark message as group event if applicable */
1747 if (unlikely(grp_evt
)) {
1748 if (msg_grp_evt(hdr
) == TIPC_WITHDRAWN
)
1749 m
->msg_flags
|= MSG_EOR
;
1750 m
->msg_flags
|= MSG_OOB
;
1754 /* Caption of data or error code/rejected data was successful */
1755 if (unlikely(flags
& MSG_PEEK
))
1758 /* Send group flow control advertisement when applicable */
1759 if (tsk
->group
&& msg_in_group(hdr
) && !grp_evt
) {
1760 skb_queue_head_init(&xmitq
);
1761 tipc_group_update_rcv_win(tsk
->group
, tsk_blocks(hlen
+ dlen
),
1762 msg_orignode(hdr
), msg_origport(hdr
),
1764 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
1767 tsk_advance_rx_queue(sk
);
1769 if (likely(!connected
))
1772 /* Send connection flow control advertisement when applicable */
1773 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1774 if (tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
)
1775 tipc_sk_send_ack(tsk
);
1778 return rc
? rc
: copy
;
1782 * tipc_recvstream - receive stream-oriented data
1783 * @m: descriptor for message info
1784 * @buflen: total size of user buffer area
1785 * @flags: receive flags
1787 * Used for SOCK_STREAM messages only. If not enough data is available
1788 * will optionally wait for more; never truncates data.
1790 * Returns size of returned message data, errno otherwise
1792 static int tipc_recvstream(struct socket
*sock
, struct msghdr
*m
,
1793 size_t buflen
, int flags
)
1795 struct sock
*sk
= sock
->sk
;
1796 struct tipc_sock
*tsk
= tipc_sk(sk
);
1797 struct sk_buff
*skb
;
1798 struct tipc_msg
*hdr
;
1799 struct tipc_skb_cb
*skb_cb
;
1800 bool peek
= flags
& MSG_PEEK
;
1801 int offset
, required
, copy
, copied
= 0;
1802 int hlen
, dlen
, err
, rc
;
1805 /* Catch invalid receive attempts */
1806 if (unlikely(!buflen
))
1811 if (unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1815 required
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buflen
);
1816 timeout
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1819 /* Look at first msg in receive queue; wait if necessary */
1820 rc
= tipc_wait_for_rcvmsg(sock
, &timeout
);
1823 skb
= skb_peek(&sk
->sk_receive_queue
);
1824 skb_cb
= TIPC_SKB_CB(skb
);
1826 dlen
= msg_data_sz(hdr
);
1827 hlen
= msg_hdr_sz(hdr
);
1828 err
= msg_errcode(hdr
);
1830 /* Discard any empty non-errored (SYN-) message */
1831 if (unlikely(!dlen
&& !err
)) {
1832 tsk_advance_rx_queue(sk
);
1836 /* Collect msg meta data, incl. error code and rejected data */
1838 tipc_sk_set_orig_addr(m
, skb
);
1839 rc
= tipc_sk_anc_data_recv(m
, hdr
, tsk
);
1844 /* Copy data if msg ok, otherwise return error/partial data */
1846 offset
= skb_cb
->bytes_read
;
1847 copy
= min_t(int, dlen
- offset
, buflen
- copied
);
1848 rc
= skb_copy_datagram_msg(skb
, hlen
+ offset
, m
, copy
);
1853 if (unlikely(offset
< dlen
)) {
1855 skb_cb
->bytes_read
= offset
;
1860 if ((err
!= TIPC_CONN_SHUTDOWN
) && !m
->msg_control
)
1869 tsk_advance_rx_queue(sk
);
1871 /* Send connection flow control advertisement when applicable */
1872 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ dlen
);
1873 if (unlikely(tsk
->rcv_unacked
>= tsk
->rcv_win
/ TIPC_ACK_RATE
))
1874 tipc_sk_send_ack(tsk
);
1876 /* Exit if all requested data or FIN/error received */
1877 if (copied
== buflen
|| err
)
1880 } while (!skb_queue_empty(&sk
->sk_receive_queue
) || copied
< required
);
1883 return copied
? copied
: rc
;
1887 * tipc_write_space - wake up thread if port congestion is released
1890 static void tipc_write_space(struct sock
*sk
)
1892 struct socket_wq
*wq
;
1895 wq
= rcu_dereference(sk
->sk_wq
);
1896 if (skwq_has_sleeper(wq
))
1897 wake_up_interruptible_sync_poll(&wq
->wait
, EPOLLOUT
|
1898 EPOLLWRNORM
| EPOLLWRBAND
);
1903 * tipc_data_ready - wake up threads to indicate messages have been received
1905 * @len: the length of messages
1907 static void tipc_data_ready(struct sock
*sk
)
1909 struct socket_wq
*wq
;
1912 wq
= rcu_dereference(sk
->sk_wq
);
1913 if (skwq_has_sleeper(wq
))
1914 wake_up_interruptible_sync_poll(&wq
->wait
, EPOLLIN
|
1915 EPOLLRDNORM
| EPOLLRDBAND
);
1919 static void tipc_sock_destruct(struct sock
*sk
)
1921 __skb_queue_purge(&sk
->sk_receive_queue
);
1924 static void tipc_sk_proto_rcv(struct sock
*sk
,
1925 struct sk_buff_head
*inputq
,
1926 struct sk_buff_head
*xmitq
)
1928 struct sk_buff
*skb
= __skb_dequeue(inputq
);
1929 struct tipc_sock
*tsk
= tipc_sk(sk
);
1930 struct tipc_msg
*hdr
= buf_msg(skb
);
1931 struct tipc_group
*grp
= tsk
->group
;
1932 bool wakeup
= false;
1934 switch (msg_user(hdr
)) {
1936 tipc_sk_conn_proto_rcv(tsk
, skb
, xmitq
);
1939 tipc_dest_del(&tsk
->cong_links
, msg_orignode(hdr
), 0);
1940 tsk
->cong_link_cnt
--;
1943 case GROUP_PROTOCOL
:
1944 tipc_group_proto_rcv(grp
, &wakeup
, hdr
, inputq
, xmitq
);
1947 tipc_group_member_evt(tsk
->group
, &wakeup
, &sk
->sk_rcvbuf
,
1948 hdr
, inputq
, xmitq
);
1955 sk
->sk_write_space(sk
);
1961 * tipc_filter_connect - Handle incoming message for a connection-based socket
1963 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1965 * Returns true if everything ok, false otherwise
1967 static bool tipc_sk_filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1969 struct sock
*sk
= &tsk
->sk
;
1970 struct net
*net
= sock_net(sk
);
1971 struct tipc_msg
*hdr
= buf_msg(skb
);
1972 u32 pport
= msg_origport(hdr
);
1973 u32 pnode
= msg_orignode(hdr
);
1975 if (unlikely(msg_mcast(hdr
)))
1978 switch (sk
->sk_state
) {
1979 case TIPC_CONNECTING
:
1980 /* Accept only ACK or NACK message */
1981 if (unlikely(!msg_connected(hdr
))) {
1982 if (pport
!= tsk_peer_port(tsk
) ||
1983 pnode
!= tsk_peer_node(tsk
))
1986 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1987 sk
->sk_err
= ECONNREFUSED
;
1988 sk
->sk_state_change(sk
);
1992 if (unlikely(msg_errcode(hdr
))) {
1993 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1994 sk
->sk_err
= ECONNREFUSED
;
1995 sk
->sk_state_change(sk
);
1999 if (unlikely(!msg_isdata(hdr
))) {
2000 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2001 sk
->sk_err
= EINVAL
;
2002 sk
->sk_state_change(sk
);
2006 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
2007 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
2009 /* If 'ACK+' message, add to socket receive queue */
2010 if (msg_data_sz(hdr
))
2013 /* If empty 'ACK-' message, wake up sleeping connect() */
2014 sk
->sk_data_ready(sk
);
2016 /* 'ACK-' message is neither accepted nor rejected: */
2017 msg_set_dest_droppable(hdr
, 1);
2021 case TIPC_DISCONNECTING
:
2024 /* Accept only SYN message */
2025 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
2028 case TIPC_ESTABLISHED
:
2029 /* Accept only connection-based messages sent by peer */
2030 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
2033 if (unlikely(msg_errcode(hdr
))) {
2034 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2035 /* Let timer expire on it's own */
2036 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
2038 sk
->sk_state_change(sk
);
2042 pr_err("Unknown sk_state %u\n", sk
->sk_state
);
2049 * rcvbuf_limit - get proper overload limit of socket receive queue
2053 * For connection oriented messages, irrespective of importance,
2054 * default queue limit is 2 MB.
2056 * For connectionless messages, queue limits are based on message
2057 * importance as follows:
2059 * TIPC_LOW_IMPORTANCE (2 MB)
2060 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2061 * TIPC_HIGH_IMPORTANCE (8 MB)
2062 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2064 * Returns overload limit according to corresponding message importance
2066 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
2068 struct tipc_sock
*tsk
= tipc_sk(sk
);
2069 struct tipc_msg
*hdr
= buf_msg(skb
);
2071 if (unlikely(msg_in_group(hdr
)))
2072 return sk
->sk_rcvbuf
;
2074 if (unlikely(!msg_connected(hdr
)))
2075 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
2077 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
2078 return sk
->sk_rcvbuf
;
2080 return FLOWCTL_MSG_LIM
;
2084 * tipc_sk_filter_rcv - validate incoming message
2086 * @skb: pointer to message.
2088 * Enqueues message on receive queue if acceptable; optionally handles
2089 * disconnect indication for a connected socket.
2091 * Called with socket lock already taken
2094 static void tipc_sk_filter_rcv(struct sock
*sk
, struct sk_buff
*skb
,
2095 struct sk_buff_head
*xmitq
)
2097 bool sk_conn
= !tipc_sk_type_connectionless(sk
);
2098 struct tipc_sock
*tsk
= tipc_sk(sk
);
2099 struct tipc_group
*grp
= tsk
->group
;
2100 struct tipc_msg
*hdr
= buf_msg(skb
);
2101 struct net
*net
= sock_net(sk
);
2102 struct sk_buff_head inputq
;
2103 int limit
, err
= TIPC_OK
;
2105 TIPC_SKB_CB(skb
)->bytes_read
= 0;
2106 __skb_queue_head_init(&inputq
);
2107 __skb_queue_tail(&inputq
, skb
);
2109 if (unlikely(!msg_isdata(hdr
)))
2110 tipc_sk_proto_rcv(sk
, &inputq
, xmitq
);
2113 tipc_group_filter_msg(grp
, &inputq
, xmitq
);
2115 /* Validate and add to receive buffer if there is space */
2116 while ((skb
= __skb_dequeue(&inputq
))) {
2118 limit
= rcvbuf_limit(sk
, skb
);
2119 if ((sk_conn
&& !tipc_sk_filter_connect(tsk
, skb
)) ||
2120 (!sk_conn
&& msg_connected(hdr
)) ||
2121 (!grp
&& msg_in_group(hdr
)))
2122 err
= TIPC_ERR_NO_PORT
;
2123 else if (sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
) {
2124 atomic_inc(&sk
->sk_drops
);
2125 err
= TIPC_ERR_OVERLOAD
;
2128 if (unlikely(err
)) {
2129 tipc_skb_reject(net
, err
, skb
, xmitq
);
2133 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
2134 skb_set_owner_r(skb
, sk
);
2135 sk
->sk_data_ready(sk
);
2140 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2144 * Caller must hold socket lock
2146 static int tipc_sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
2148 unsigned int before
= sk_rmem_alloc_get(sk
);
2149 struct sk_buff_head xmitq
;
2152 __skb_queue_head_init(&xmitq
);
2154 tipc_sk_filter_rcv(sk
, skb
, &xmitq
);
2155 added
= sk_rmem_alloc_get(sk
) - before
;
2156 atomic_add(added
, &tipc_sk(sk
)->dupl_rcvcnt
);
2158 /* Send pending response/rejected messages, if any */
2159 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2164 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2165 * inputq and try adding them to socket or backlog queue
2166 * @inputq: list of incoming buffers with potentially different destinations
2167 * @sk: socket where the buffers should be enqueued
2168 * @dport: port number for the socket
2170 * Caller must hold socket lock
2172 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
2173 u32 dport
, struct sk_buff_head
*xmitq
)
2175 unsigned long time_limit
= jiffies
+ 2;
2176 struct sk_buff
*skb
;
2181 while (skb_queue_len(inputq
)) {
2182 if (unlikely(time_after_eq(jiffies
, time_limit
)))
2185 skb
= tipc_skb_dequeue(inputq
, dport
);
2189 /* Add message directly to receive queue if possible */
2190 if (!sock_owned_by_user(sk
)) {
2191 tipc_sk_filter_rcv(sk
, skb
, xmitq
);
2195 /* Try backlog, compensating for double-counted bytes */
2196 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
2197 if (!sk
->sk_backlog
.len
)
2198 atomic_set(dcnt
, 0);
2199 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
2200 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
2203 /* Overload => reject message back to sender */
2204 onode
= tipc_own_addr(sock_net(sk
));
2205 atomic_inc(&sk
->sk_drops
);
2206 if (tipc_msg_reverse(onode
, &skb
, TIPC_ERR_OVERLOAD
))
2207 __skb_queue_tail(xmitq
, skb
);
2213 * tipc_sk_rcv - handle a chain of incoming buffers
2214 * @inputq: buffer list containing the buffers
2215 * Consumes all buffers in list until inputq is empty
2216 * Note: may be called in multiple threads referring to the same queue
2218 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
2220 struct sk_buff_head xmitq
;
2221 u32 dnode
, dport
= 0;
2223 struct tipc_sock
*tsk
;
2225 struct sk_buff
*skb
;
2227 __skb_queue_head_init(&xmitq
);
2228 while (skb_queue_len(inputq
)) {
2229 dport
= tipc_skb_peek_port(inputq
, dport
);
2230 tsk
= tipc_sk_lookup(net
, dport
);
2234 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
2235 tipc_sk_enqueue(inputq
, sk
, dport
, &xmitq
);
2236 spin_unlock_bh(&sk
->sk_lock
.slock
);
2238 /* Send pending response/rejected messages, if any */
2239 tipc_node_distr_xmit(sock_net(sk
), &xmitq
);
2243 /* No destination socket => dequeue skb if still there */
2244 skb
= tipc_skb_dequeue(inputq
, dport
);
2248 /* Try secondary lookup if unresolved named message */
2249 err
= TIPC_ERR_NO_PORT
;
2250 if (tipc_msg_lookup_dest(net
, skb
, &err
))
2253 /* Prepare for message rejection */
2254 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
2257 dnode
= msg_destnode(buf_msg(skb
));
2258 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
2262 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
2264 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
2265 struct sock
*sk
= sock
->sk
;
2269 int err
= sock_error(sk
);
2274 if (signal_pending(current
))
2275 return sock_intr_errno(*timeo_p
);
2277 add_wait_queue(sk_sleep(sk
), &wait
);
2278 done
= sk_wait_event(sk
, timeo_p
,
2279 sk
->sk_state
!= TIPC_CONNECTING
, &wait
);
2280 remove_wait_queue(sk_sleep(sk
), &wait
);
2286 * tipc_connect - establish a connection to another TIPC port
2287 * @sock: socket structure
2288 * @dest: socket address for destination port
2289 * @destlen: size of socket address data structure
2290 * @flags: file-related flags associated with socket
2292 * Returns 0 on success, errno otherwise
2294 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
2295 int destlen
, int flags
)
2297 struct sock
*sk
= sock
->sk
;
2298 struct tipc_sock
*tsk
= tipc_sk(sk
);
2299 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
2300 struct msghdr m
= {NULL
,};
2301 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
2305 if (destlen
!= sizeof(struct sockaddr_tipc
))
2315 if (dst
->family
== AF_UNSPEC
) {
2316 memset(&tsk
->peer
, 0, sizeof(struct sockaddr_tipc
));
2317 if (!tipc_sk_type_connectionless(sk
))
2320 } else if (dst
->family
!= AF_TIPC
) {
2323 if (dst
->addrtype
!= TIPC_ADDR_ID
&& dst
->addrtype
!= TIPC_ADDR_NAME
)
2328 /* DGRAM/RDM connect(), just save the destaddr */
2329 if (tipc_sk_type_connectionless(sk
)) {
2330 memcpy(&tsk
->peer
, dest
, destlen
);
2334 previous
= sk
->sk_state
;
2336 switch (sk
->sk_state
) {
2338 /* Send a 'SYN-' to destination */
2340 m
.msg_namelen
= destlen
;
2342 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2343 * indicate send_msg() is never blocked.
2346 m
.msg_flags
= MSG_DONTWAIT
;
2348 res
= __tipc_sendmsg(sock
, &m
, 0);
2349 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
2352 /* Just entered TIPC_CONNECTING state; the only
2353 * difference is that return value in non-blocking
2354 * case is EINPROGRESS, rather than EALREADY.
2358 case TIPC_CONNECTING
:
2360 if (previous
== TIPC_CONNECTING
)
2364 timeout
= msecs_to_jiffies(timeout
);
2365 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2366 res
= tipc_wait_for_connect(sock
, &timeout
);
2368 case TIPC_ESTABLISHED
:
2381 * tipc_listen - allow socket to listen for incoming connections
2382 * @sock: socket structure
2385 * Returns 0 on success, errno otherwise
2387 static int tipc_listen(struct socket
*sock
, int len
)
2389 struct sock
*sk
= sock
->sk
;
2393 res
= tipc_set_sk_state(sk
, TIPC_LISTEN
);
2399 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
2401 struct sock
*sk
= sock
->sk
;
2405 /* True wake-one mechanism for incoming connections: only
2406 * one process gets woken up, not the 'whole herd'.
2407 * Since we do not 'race & poll' for established sockets
2408 * anymore, the common case will execute the loop only once.
2411 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2412 TASK_INTERRUPTIBLE
);
2413 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2415 timeo
= schedule_timeout(timeo
);
2419 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2424 err
= sock_intr_errno(timeo
);
2425 if (signal_pending(current
))
2428 finish_wait(sk_sleep(sk
), &wait
);
2433 * tipc_accept - wait for connection request
2434 * @sock: listening socket
2435 * @newsock: new socket that is to be connected
2436 * @flags: file-related flags associated with socket
2438 * Returns 0 on success, errno otherwise
2440 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
,
2443 struct sock
*new_sk
, *sk
= sock
->sk
;
2444 struct sk_buff
*buf
;
2445 struct tipc_sock
*new_tsock
;
2446 struct tipc_msg
*msg
;
2452 if (sk
->sk_state
!= TIPC_LISTEN
) {
2456 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2457 res
= tipc_wait_for_accept(sock
, timeo
);
2461 buf
= skb_peek(&sk
->sk_receive_queue
);
2463 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, kern
);
2466 security_sk_clone(sock
->sk
, new_sock
->sk
);
2468 new_sk
= new_sock
->sk
;
2469 new_tsock
= tipc_sk(new_sk
);
2472 /* we lock on new_sk; but lockdep sees the lock on sk */
2473 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2476 * Reject any stray messages received by new socket
2477 * before the socket lock was taken (very, very unlikely)
2479 tsk_rej_rx_queue(new_sk
);
2481 /* Connect new socket to it's peer */
2482 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2484 tsk_set_importance(new_tsock
, msg_importance(msg
));
2485 if (msg_named(msg
)) {
2486 new_tsock
->conn_type
= msg_nametype(msg
);
2487 new_tsock
->conn_instance
= msg_nameinst(msg
);
2491 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2492 * Respond to 'SYN+' by queuing it on new socket.
2494 if (!msg_data_sz(msg
)) {
2495 struct msghdr m
= {NULL
,};
2497 tsk_advance_rx_queue(sk
);
2498 __tipc_sendstream(new_sock
, &m
, 0);
2500 __skb_dequeue(&sk
->sk_receive_queue
);
2501 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2502 skb_set_owner_r(buf
, new_sk
);
2504 release_sock(new_sk
);
2511 * tipc_shutdown - shutdown socket connection
2512 * @sock: socket structure
2513 * @how: direction to close (must be SHUT_RDWR)
2515 * Terminates connection (if necessary), then purges socket's receive queue.
2517 * Returns 0 on success, errno otherwise
2519 static int tipc_shutdown(struct socket
*sock
, int how
)
2521 struct sock
*sk
= sock
->sk
;
2524 if (how
!= SHUT_RDWR
)
2529 __tipc_shutdown(sock
, TIPC_CONN_SHUTDOWN
);
2530 sk
->sk_shutdown
= SEND_SHUTDOWN
;
2532 if (sk
->sk_state
== TIPC_DISCONNECTING
) {
2533 /* Discard any unreceived messages */
2534 __skb_queue_purge(&sk
->sk_receive_queue
);
2536 /* Wake up anyone sleeping in poll */
2537 sk
->sk_state_change(sk
);
2547 static void tipc_sk_timeout(struct timer_list
*t
)
2549 struct sock
*sk
= from_timer(sk
, t
, sk_timer
);
2550 struct tipc_sock
*tsk
= tipc_sk(sk
);
2551 u32 peer_port
= tsk_peer_port(tsk
);
2552 u32 peer_node
= tsk_peer_node(tsk
);
2553 u32 own_node
= tsk_own_node(tsk
);
2554 u32 own_port
= tsk
->portid
;
2555 struct net
*net
= sock_net(sk
);
2556 struct sk_buff
*skb
= NULL
;
2559 if (!tipc_sk_connected(sk
))
2562 /* Try again later if socket is busy */
2563 if (sock_owned_by_user(sk
)) {
2564 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ HZ
/ 20);
2568 if (tsk
->probe_unacked
) {
2569 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2570 tipc_node_remove_conn(net
, peer_node
, peer_port
);
2571 sk
->sk_state_change(sk
);
2574 /* Send new probe */
2575 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
, INT_H_SIZE
, 0,
2576 peer_node
, own_node
, peer_port
, own_port
,
2578 tsk
->probe_unacked
= true;
2579 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTV
);
2583 tipc_node_xmit_skb(net
, skb
, peer_node
, own_port
);
2587 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2588 struct tipc_name_seq
const *seq
)
2590 struct sock
*sk
= &tsk
->sk
;
2591 struct net
*net
= sock_net(sk
);
2592 struct publication
*publ
;
2595 if (scope
!= TIPC_NODE_SCOPE
)
2596 scope
= TIPC_CLUSTER_SCOPE
;
2598 if (tipc_sk_connected(sk
))
2600 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2601 if (key
== tsk
->portid
)
2604 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2605 scope
, tsk
->portid
, key
);
2606 if (unlikely(!publ
))
2609 list_add(&publ
->binding_sock
, &tsk
->publications
);
2615 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2616 struct tipc_name_seq
const *seq
)
2618 struct net
*net
= sock_net(&tsk
->sk
);
2619 struct publication
*publ
;
2620 struct publication
*safe
;
2623 if (scope
!= TIPC_NODE_SCOPE
)
2624 scope
= TIPC_CLUSTER_SCOPE
;
2626 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, binding_sock
) {
2628 if (publ
->scope
!= scope
)
2630 if (publ
->type
!= seq
->type
)
2632 if (publ
->lower
!= seq
->lower
)
2634 if (publ
->upper
!= seq
->upper
)
2636 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2637 publ
->upper
, publ
->key
);
2641 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2642 publ
->upper
, publ
->key
);
2645 if (list_empty(&tsk
->publications
))
2650 /* tipc_sk_reinit: set non-zero address in all existing sockets
2651 * when we go from standalone to network mode.
2653 void tipc_sk_reinit(struct net
*net
)
2655 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2656 struct rhashtable_iter iter
;
2657 struct tipc_sock
*tsk
;
2658 struct tipc_msg
*msg
;
2660 rhashtable_walk_enter(&tn
->sk_rht
, &iter
);
2663 rhashtable_walk_start(&iter
);
2665 while ((tsk
= rhashtable_walk_next(&iter
)) && !IS_ERR(tsk
)) {
2666 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2668 msg_set_prevnode(msg
, tipc_own_addr(net
));
2669 msg_set_orignode(msg
, tipc_own_addr(net
));
2670 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2673 rhashtable_walk_stop(&iter
);
2674 } while (tsk
== ERR_PTR(-EAGAIN
));
2677 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2679 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2680 struct tipc_sock
*tsk
;
2683 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2685 sock_hold(&tsk
->sk
);
2691 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2693 struct sock
*sk
= &tsk
->sk
;
2694 struct net
*net
= sock_net(sk
);
2695 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2696 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2697 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2699 while (remaining
--) {
2701 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2702 portid
= TIPC_MIN_PORT
;
2703 tsk
->portid
= portid
;
2704 sock_hold(&tsk
->sk
);
2705 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2714 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2716 struct sock
*sk
= &tsk
->sk
;
2717 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2719 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2720 WARN_ON(refcount_read(&sk
->sk_refcnt
) == 1);
2725 static const struct rhashtable_params tsk_rht_params
= {
2727 .head_offset
= offsetof(struct tipc_sock
, node
),
2728 .key_offset
= offsetof(struct tipc_sock
, portid
),
2729 .key_len
= sizeof(u32
), /* portid */
2730 .max_size
= 1048576,
2732 .automatic_shrinking
= true,
2735 int tipc_sk_rht_init(struct net
*net
)
2737 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2739 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2742 void tipc_sk_rht_destroy(struct net
*net
)
2744 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2746 /* Wait for socket readers to complete */
2749 rhashtable_destroy(&tn
->sk_rht
);
2752 static int tipc_sk_join(struct tipc_sock
*tsk
, struct tipc_group_req
*mreq
)
2754 struct net
*net
= sock_net(&tsk
->sk
);
2755 struct tipc_group
*grp
= tsk
->group
;
2756 struct tipc_msg
*hdr
= &tsk
->phdr
;
2757 struct tipc_name_seq seq
;
2760 if (mreq
->type
< TIPC_RESERVED_TYPES
)
2762 if (mreq
->scope
> TIPC_NODE_SCOPE
)
2766 grp
= tipc_group_create(net
, tsk
->portid
, mreq
, &tsk
->group_is_open
);
2770 msg_set_lookup_scope(hdr
, mreq
->scope
);
2771 msg_set_nametype(hdr
, mreq
->type
);
2772 msg_set_dest_droppable(hdr
, true);
2773 seq
.type
= mreq
->type
;
2774 seq
.lower
= mreq
->instance
;
2775 seq
.upper
= seq
.lower
;
2776 tipc_nametbl_build_group(net
, grp
, mreq
->type
, mreq
->scope
);
2777 rc
= tipc_sk_publish(tsk
, mreq
->scope
, &seq
);
2779 tipc_group_delete(net
, grp
);
2783 /* Eliminate any risk that a broadcast overtakes sent JOINs */
2784 tsk
->mc_method
.rcast
= true;
2785 tsk
->mc_method
.mandatory
= true;
2786 tipc_group_join(net
, grp
, &tsk
->sk
.sk_rcvbuf
);
2790 static int tipc_sk_leave(struct tipc_sock
*tsk
)
2792 struct net
*net
= sock_net(&tsk
->sk
);
2793 struct tipc_group
*grp
= tsk
->group
;
2794 struct tipc_name_seq seq
;
2799 tipc_group_self(grp
, &seq
, &scope
);
2800 tipc_group_delete(net
, grp
);
2802 tipc_sk_withdraw(tsk
, scope
, &seq
);
2807 * tipc_setsockopt - set socket option
2808 * @sock: socket structure
2809 * @lvl: option level
2810 * @opt: option identifier
2811 * @ov: pointer to new option value
2812 * @ol: length of option value
2814 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2815 * (to ease compatibility).
2817 * Returns 0 on success, errno otherwise
2819 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2820 char __user
*ov
, unsigned int ol
)
2822 struct sock
*sk
= sock
->sk
;
2823 struct tipc_sock
*tsk
= tipc_sk(sk
);
2824 struct tipc_group_req mreq
;
2828 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2830 if (lvl
!= SOL_TIPC
)
2831 return -ENOPROTOOPT
;
2834 case TIPC_IMPORTANCE
:
2835 case TIPC_SRC_DROPPABLE
:
2836 case TIPC_DEST_DROPPABLE
:
2837 case TIPC_CONN_TIMEOUT
:
2838 if (ol
< sizeof(value
))
2840 if (get_user(value
, (u32 __user
*)ov
))
2843 case TIPC_GROUP_JOIN
:
2844 if (ol
< sizeof(mreq
))
2846 if (copy_from_user(&mreq
, ov
, sizeof(mreq
)))
2857 case TIPC_IMPORTANCE
:
2858 res
= tsk_set_importance(tsk
, value
);
2860 case TIPC_SRC_DROPPABLE
:
2861 if (sock
->type
!= SOCK_STREAM
)
2862 tsk_set_unreliable(tsk
, value
);
2866 case TIPC_DEST_DROPPABLE
:
2867 tsk_set_unreturnable(tsk
, value
);
2869 case TIPC_CONN_TIMEOUT
:
2870 tipc_sk(sk
)->conn_timeout
= value
;
2872 case TIPC_MCAST_BROADCAST
:
2873 tsk
->mc_method
.rcast
= false;
2874 tsk
->mc_method
.mandatory
= true;
2876 case TIPC_MCAST_REPLICAST
:
2877 tsk
->mc_method
.rcast
= true;
2878 tsk
->mc_method
.mandatory
= true;
2880 case TIPC_GROUP_JOIN
:
2881 res
= tipc_sk_join(tsk
, &mreq
);
2883 case TIPC_GROUP_LEAVE
:
2884 res
= tipc_sk_leave(tsk
);
2896 * tipc_getsockopt - get socket option
2897 * @sock: socket structure
2898 * @lvl: option level
2899 * @opt: option identifier
2900 * @ov: receptacle for option value
2901 * @ol: receptacle for length of option value
2903 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2904 * (to ease compatibility).
2906 * Returns 0 on success, errno otherwise
2908 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2909 char __user
*ov
, int __user
*ol
)
2911 struct sock
*sk
= sock
->sk
;
2912 struct tipc_sock
*tsk
= tipc_sk(sk
);
2913 struct tipc_name_seq seq
;
2918 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2919 return put_user(0, ol
);
2920 if (lvl
!= SOL_TIPC
)
2921 return -ENOPROTOOPT
;
2922 res
= get_user(len
, ol
);
2929 case TIPC_IMPORTANCE
:
2930 value
= tsk_importance(tsk
);
2932 case TIPC_SRC_DROPPABLE
:
2933 value
= tsk_unreliable(tsk
);
2935 case TIPC_DEST_DROPPABLE
:
2936 value
= tsk_unreturnable(tsk
);
2938 case TIPC_CONN_TIMEOUT
:
2939 value
= tsk
->conn_timeout
;
2940 /* no need to set "res", since already 0 at this point */
2942 case TIPC_NODE_RECVQ_DEPTH
:
2943 value
= 0; /* was tipc_queue_size, now obsolete */
2945 case TIPC_SOCK_RECVQ_DEPTH
:
2946 value
= skb_queue_len(&sk
->sk_receive_queue
);
2948 case TIPC_GROUP_JOIN
:
2951 tipc_group_self(tsk
->group
, &seq
, &scope
);
2961 return res
; /* "get" failed */
2963 if (len
< sizeof(value
))
2966 if (copy_to_user(ov
, &value
, sizeof(value
)))
2969 return put_user(sizeof(value
), ol
);
2972 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2974 struct net
*net
= sock_net(sock
->sk
);
2975 struct tipc_sioc_nodeid_req nr
= {0};
2976 struct tipc_sioc_ln_req lnr
;
2977 void __user
*argp
= (void __user
*)arg
;
2980 case SIOCGETLINKNAME
:
2981 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2983 if (!tipc_node_get_linkname(net
,
2984 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2985 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2986 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2990 return -EADDRNOTAVAIL
;
2992 if (copy_from_user(&nr
, argp
, sizeof(nr
)))
2994 if (!tipc_node_get_id(net
, nr
.peer
, nr
.node_id
))
2995 return -EADDRNOTAVAIL
;
2996 if (copy_to_user(argp
, &nr
, sizeof(nr
)))
3000 return -ENOIOCTLCMD
;
3004 static int tipc_socketpair(struct socket
*sock1
, struct socket
*sock2
)
3006 struct tipc_sock
*tsk2
= tipc_sk(sock2
->sk
);
3007 struct tipc_sock
*tsk1
= tipc_sk(sock1
->sk
);
3008 u32 onode
= tipc_own_addr(sock_net(sock1
->sk
));
3010 tsk1
->peer
.family
= AF_TIPC
;
3011 tsk1
->peer
.addrtype
= TIPC_ADDR_ID
;
3012 tsk1
->peer
.scope
= TIPC_NODE_SCOPE
;
3013 tsk1
->peer
.addr
.id
.ref
= tsk2
->portid
;
3014 tsk1
->peer
.addr
.id
.node
= onode
;
3015 tsk2
->peer
.family
= AF_TIPC
;
3016 tsk2
->peer
.addrtype
= TIPC_ADDR_ID
;
3017 tsk2
->peer
.scope
= TIPC_NODE_SCOPE
;
3018 tsk2
->peer
.addr
.id
.ref
= tsk1
->portid
;
3019 tsk2
->peer
.addr
.id
.node
= onode
;
3021 tipc_sk_finish_conn(tsk1
, tsk2
->portid
, onode
);
3022 tipc_sk_finish_conn(tsk2
, tsk1
->portid
, onode
);
3026 /* Protocol switches for the various types of TIPC sockets */
3028 static const struct proto_ops msg_ops
= {
3029 .owner
= THIS_MODULE
,
3031 .release
= tipc_release
,
3033 .connect
= tipc_connect
,
3034 .socketpair
= tipc_socketpair
,
3035 .accept
= sock_no_accept
,
3036 .getname
= tipc_getname
,
3038 .ioctl
= tipc_ioctl
,
3039 .listen
= sock_no_listen
,
3040 .shutdown
= tipc_shutdown
,
3041 .setsockopt
= tipc_setsockopt
,
3042 .getsockopt
= tipc_getsockopt
,
3043 .sendmsg
= tipc_sendmsg
,
3044 .recvmsg
= tipc_recvmsg
,
3045 .mmap
= sock_no_mmap
,
3046 .sendpage
= sock_no_sendpage
3049 static const struct proto_ops packet_ops
= {
3050 .owner
= THIS_MODULE
,
3052 .release
= tipc_release
,
3054 .connect
= tipc_connect
,
3055 .socketpair
= tipc_socketpair
,
3056 .accept
= tipc_accept
,
3057 .getname
= tipc_getname
,
3059 .ioctl
= tipc_ioctl
,
3060 .listen
= tipc_listen
,
3061 .shutdown
= tipc_shutdown
,
3062 .setsockopt
= tipc_setsockopt
,
3063 .getsockopt
= tipc_getsockopt
,
3064 .sendmsg
= tipc_send_packet
,
3065 .recvmsg
= tipc_recvmsg
,
3066 .mmap
= sock_no_mmap
,
3067 .sendpage
= sock_no_sendpage
3070 static const struct proto_ops stream_ops
= {
3071 .owner
= THIS_MODULE
,
3073 .release
= tipc_release
,
3075 .connect
= tipc_connect
,
3076 .socketpair
= tipc_socketpair
,
3077 .accept
= tipc_accept
,
3078 .getname
= tipc_getname
,
3080 .ioctl
= tipc_ioctl
,
3081 .listen
= tipc_listen
,
3082 .shutdown
= tipc_shutdown
,
3083 .setsockopt
= tipc_setsockopt
,
3084 .getsockopt
= tipc_getsockopt
,
3085 .sendmsg
= tipc_sendstream
,
3086 .recvmsg
= tipc_recvstream
,
3087 .mmap
= sock_no_mmap
,
3088 .sendpage
= sock_no_sendpage
3091 static const struct net_proto_family tipc_family_ops
= {
3092 .owner
= THIS_MODULE
,
3094 .create
= tipc_sk_create
3097 static struct proto tipc_proto
= {
3099 .owner
= THIS_MODULE
,
3100 .obj_size
= sizeof(struct tipc_sock
),
3101 .sysctl_rmem
= sysctl_tipc_rmem
3105 * tipc_socket_init - initialize TIPC socket interface
3107 * Returns 0 on success, errno otherwise
3109 int tipc_socket_init(void)
3113 res
= proto_register(&tipc_proto
, 1);
3115 pr_err("Failed to register TIPC protocol type\n");
3119 res
= sock_register(&tipc_family_ops
);
3121 pr_err("Failed to register TIPC socket type\n");
3122 proto_unregister(&tipc_proto
);
3130 * tipc_socket_stop - stop TIPC socket interface
3132 void tipc_socket_stop(void)
3134 sock_unregister(tipc_family_ops
.family
);
3135 proto_unregister(&tipc_proto
);
3138 /* Caller should hold socket lock for the passed tipc socket. */
3139 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
3143 struct nlattr
*nest
;
3145 peer_node
= tsk_peer_node(tsk
);
3146 peer_port
= tsk_peer_port(tsk
);
3148 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
3150 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
3152 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
3155 if (tsk
->conn_type
!= 0) {
3156 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
3158 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
3160 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
3163 nla_nest_end(skb
, nest
);
3168 nla_nest_cancel(skb
, nest
);
3173 static int __tipc_nl_add_sk_info(struct sk_buff
*skb
, struct tipc_sock
3176 struct net
*net
= sock_net(skb
->sk
);
3177 struct sock
*sk
= &tsk
->sk
;
3179 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
) ||
3180 nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tipc_own_addr(net
)))
3183 if (tipc_sk_connected(sk
)) {
3184 if (__tipc_nl_add_sk_con(skb
, tsk
))
3186 } else if (!list_empty(&tsk
->publications
)) {
3187 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
3193 /* Caller should hold socket lock for the passed tipc socket. */
3194 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3195 struct tipc_sock
*tsk
)
3197 struct nlattr
*attrs
;
3200 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3201 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
3205 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
3207 goto genlmsg_cancel
;
3209 if (__tipc_nl_add_sk_info(skb
, tsk
))
3210 goto attr_msg_cancel
;
3212 nla_nest_end(skb
, attrs
);
3213 genlmsg_end(skb
, hdr
);
3218 nla_nest_cancel(skb
, attrs
);
3220 genlmsg_cancel(skb
, hdr
);
3225 int tipc_nl_sk_walk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3226 int (*skb_handler
)(struct sk_buff
*skb
,
3227 struct netlink_callback
*cb
,
3228 struct tipc_sock
*tsk
))
3230 struct net
*net
= sock_net(skb
->sk
);
3231 struct tipc_net
*tn
= tipc_net(net
);
3232 const struct bucket_table
*tbl
;
3233 u32 prev_portid
= cb
->args
[1];
3234 u32 tbl_id
= cb
->args
[0];
3235 struct rhash_head
*pos
;
3236 struct tipc_sock
*tsk
;
3240 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
3241 for (; tbl_id
< tbl
->size
; tbl_id
++) {
3242 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
3243 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
3244 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
3245 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3249 err
= skb_handler(skb
, cb
, tsk
);
3251 prev_portid
= tsk
->portid
;
3252 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3257 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
3262 cb
->args
[0] = tbl_id
;
3263 cb
->args
[1] = prev_portid
;
3267 EXPORT_SYMBOL(tipc_nl_sk_walk
);
3269 int tipc_sk_fill_sock_diag(struct sk_buff
*skb
, struct netlink_callback
*cb
,
3270 struct tipc_sock
*tsk
, u32 sk_filter_state
,
3271 u64 (*tipc_diag_gen_cookie
)(struct sock
*sk
))
3273 struct sock
*sk
= &tsk
->sk
;
3274 struct nlattr
*attrs
;
3275 struct nlattr
*stat
;
3277 /*filter response w.r.t sk_state*/
3278 if (!(sk_filter_state
& (1 << sk
->sk_state
)))
3281 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
3285 if (__tipc_nl_add_sk_info(skb
, tsk
))
3286 goto attr_msg_cancel
;
3288 if (nla_put_u32(skb
, TIPC_NLA_SOCK_TYPE
, (u32
)sk
->sk_type
) ||
3289 nla_put_u32(skb
, TIPC_NLA_SOCK_TIPC_STATE
, (u32
)sk
->sk_state
) ||
3290 nla_put_u32(skb
, TIPC_NLA_SOCK_INO
, sock_i_ino(sk
)) ||
3291 nla_put_u32(skb
, TIPC_NLA_SOCK_UID
,
3292 from_kuid_munged(sk_user_ns(NETLINK_CB(cb
->skb
).sk
),
3294 nla_put_u64_64bit(skb
, TIPC_NLA_SOCK_COOKIE
,
3295 tipc_diag_gen_cookie(sk
),
3297 goto attr_msg_cancel
;
3299 stat
= nla_nest_start(skb
, TIPC_NLA_SOCK_STAT
);
3301 goto attr_msg_cancel
;
3303 if (nla_put_u32(skb
, TIPC_NLA_SOCK_STAT_RCVQ
,
3304 skb_queue_len(&sk
->sk_receive_queue
)) ||
3305 nla_put_u32(skb
, TIPC_NLA_SOCK_STAT_SENDQ
,
3306 skb_queue_len(&sk
->sk_write_queue
)) ||
3307 nla_put_u32(skb
, TIPC_NLA_SOCK_STAT_DROP
,
3308 atomic_read(&sk
->sk_drops
)))
3309 goto stat_msg_cancel
;
3311 if (tsk
->cong_link_cnt
&&
3312 nla_put_flag(skb
, TIPC_NLA_SOCK_STAT_LINK_CONG
))
3313 goto stat_msg_cancel
;
3315 if (tsk_conn_cong(tsk
) &&
3316 nla_put_flag(skb
, TIPC_NLA_SOCK_STAT_CONN_CONG
))
3317 goto stat_msg_cancel
;
3319 nla_nest_end(skb
, stat
);
3322 if (tipc_group_fill_sock_diag(tsk
->group
, skb
))
3323 goto stat_msg_cancel
;
3325 nla_nest_end(skb
, attrs
);
3330 nla_nest_cancel(skb
, stat
);
3332 nla_nest_cancel(skb
, attrs
);
3336 EXPORT_SYMBOL(tipc_sk_fill_sock_diag
);
3338 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3340 return tipc_nl_sk_walk(skb
, cb
, __tipc_nl_add_sk
);
3343 /* Caller should hold socket lock for the passed tipc socket. */
3344 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
3345 struct netlink_callback
*cb
,
3346 struct publication
*publ
)
3349 struct nlattr
*attrs
;
3351 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3352 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
3356 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
3358 goto genlmsg_cancel
;
3360 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
3361 goto attr_msg_cancel
;
3362 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
3363 goto attr_msg_cancel
;
3364 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
3365 goto attr_msg_cancel
;
3366 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
3367 goto attr_msg_cancel
;
3369 nla_nest_end(skb
, attrs
);
3370 genlmsg_end(skb
, hdr
);
3375 nla_nest_cancel(skb
, attrs
);
3377 genlmsg_cancel(skb
, hdr
);
3382 /* Caller should hold socket lock for the passed tipc socket. */
3383 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
3384 struct netlink_callback
*cb
,
3385 struct tipc_sock
*tsk
, u32
*last_publ
)
3388 struct publication
*p
;
3391 list_for_each_entry(p
, &tsk
->publications
, binding_sock
) {
3392 if (p
->key
== *last_publ
)
3395 if (p
->key
!= *last_publ
) {
3396 /* We never set seq or call nl_dump_check_consistent()
3397 * this means that setting prev_seq here will cause the
3398 * consistence check to fail in the netlink callback
3399 * handler. Resulting in the last NLMSG_DONE message
3400 * having the NLM_F_DUMP_INTR flag set.
3407 p
= list_first_entry(&tsk
->publications
, struct publication
,
3411 list_for_each_entry_from(p
, &tsk
->publications
, binding_sock
) {
3412 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
3414 *last_publ
= p
->key
;
3423 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3426 u32 tsk_portid
= cb
->args
[0];
3427 u32 last_publ
= cb
->args
[1];
3428 u32 done
= cb
->args
[2];
3429 struct net
*net
= sock_net(skb
->sk
);
3430 struct tipc_sock
*tsk
;
3433 struct nlattr
**attrs
;
3434 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
3436 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
3440 if (!attrs
[TIPC_NLA_SOCK
])
3443 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
3444 attrs
[TIPC_NLA_SOCK
],
3445 tipc_nl_sock_policy
, NULL
);
3449 if (!sock
[TIPC_NLA_SOCK_REF
])
3452 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
3458 tsk
= tipc_sk_lookup(net
, tsk_portid
);
3462 lock_sock(&tsk
->sk
);
3463 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
3466 release_sock(&tsk
->sk
);
3469 cb
->args
[0] = tsk_portid
;
3470 cb
->args
[1] = last_publ
;