2 * net/tipc/socket.c: TIPC socket API
4 * Copyright (c) 2001-2007, 2012-2016, 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>
39 #include "name_table.h"
42 #include "name_distr.h"
47 #define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
48 #define CONN_PROBING_INTERVAL msecs_to_jiffies(3600000) /* [ms] => 1 h */
49 #define TIPC_FWD_MSG 1
50 #define TIPC_MAX_PORT 0xffffffff
51 #define TIPC_MIN_PORT 1
54 TIPC_LISTEN
= TCP_LISTEN
,
55 TIPC_ESTABLISHED
= TCP_ESTABLISHED
,
56 TIPC_OPEN
= TCP_CLOSE
,
57 TIPC_DISCONNECTING
= TCP_CLOSE_WAIT
,
58 TIPC_CONNECTING
= TCP_SYN_SENT
,
62 * struct tipc_sock - TIPC socket structure
63 * @sk: socket - interacts with 'port' and with user via the socket API
64 * @conn_type: TIPC type used when connection was established
65 * @conn_instance: TIPC instance used when connection was established
66 * @published: non-zero if port has one or more associated names
67 * @max_pkt: maximum packet size "hint" used when building messages sent by port
68 * @portid: unique port identity in TIPC socket hash table
69 * @phdr: preformatted message header used when sending messages
70 * #cong_links: list of congested links
71 * @publications: list of publications for port
72 * @blocking_link: address of the congested link we are currently sleeping on
73 * @pub_count: total # of publications port has made during its lifetime
75 * @conn_timeout: the time we can wait for an unresponded setup request
76 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
77 * @cong_link_cnt: number of congested links
78 * @sent_unacked: # messages sent by socket, and not yet acked by peer
79 * @rcv_unacked: # messages read by user, but not yet acked back to peer
80 * @peer: 'connected' peer for dgram/rdm
81 * @node: hash table node
82 * @mc_method: cookie for use between socket and broadcast layer
83 * @rcu: rcu struct for tipc_sock
93 struct list_head cong_links
;
94 struct list_head publications
;
105 struct sockaddr_tipc peer
;
106 struct rhash_head node
;
107 struct tipc_mc_method mc_method
;
111 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
);
112 static void tipc_data_ready(struct sock
*sk
);
113 static void tipc_write_space(struct sock
*sk
);
114 static void tipc_sock_destruct(struct sock
*sk
);
115 static int tipc_release(struct socket
*sock
);
116 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
);
117 static void tipc_sk_timeout(unsigned long data
);
118 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
119 struct tipc_name_seq
const *seq
);
120 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
121 struct tipc_name_seq
const *seq
);
122 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
);
123 static int tipc_sk_insert(struct tipc_sock
*tsk
);
124 static void tipc_sk_remove(struct tipc_sock
*tsk
);
125 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
126 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dsz
);
128 static const struct proto_ops packet_ops
;
129 static const struct proto_ops stream_ops
;
130 static const struct proto_ops msg_ops
;
131 static struct proto tipc_proto
;
132 static const struct rhashtable_params tsk_rht_params
;
134 static u32
tsk_own_node(struct tipc_sock
*tsk
)
136 return msg_prevnode(&tsk
->phdr
);
139 static u32
tsk_peer_node(struct tipc_sock
*tsk
)
141 return msg_destnode(&tsk
->phdr
);
144 static u32
tsk_peer_port(struct tipc_sock
*tsk
)
146 return msg_destport(&tsk
->phdr
);
149 static bool tsk_unreliable(struct tipc_sock
*tsk
)
151 return msg_src_droppable(&tsk
->phdr
) != 0;
154 static void tsk_set_unreliable(struct tipc_sock
*tsk
, bool unreliable
)
156 msg_set_src_droppable(&tsk
->phdr
, unreliable
? 1 : 0);
159 static bool tsk_unreturnable(struct tipc_sock
*tsk
)
161 return msg_dest_droppable(&tsk
->phdr
) != 0;
164 static void tsk_set_unreturnable(struct tipc_sock
*tsk
, bool unreturnable
)
166 msg_set_dest_droppable(&tsk
->phdr
, unreturnable
? 1 : 0);
169 static int tsk_importance(struct tipc_sock
*tsk
)
171 return msg_importance(&tsk
->phdr
);
174 static int tsk_set_importance(struct tipc_sock
*tsk
, int imp
)
176 if (imp
> TIPC_CRITICAL_IMPORTANCE
)
178 msg_set_importance(&tsk
->phdr
, (u32
)imp
);
182 static struct tipc_sock
*tipc_sk(const struct sock
*sk
)
184 return container_of(sk
, struct tipc_sock
, sk
);
187 static bool tsk_conn_cong(struct tipc_sock
*tsk
)
189 return tsk
->snt_unacked
> tsk
->snd_win
;
192 /* tsk_blocks(): translate a buffer size in bytes to number of
193 * advertisable blocks, taking into account the ratio truesize(len)/len
194 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
196 static u16
tsk_adv_blocks(int len
)
198 return len
/ FLOWCTL_BLK_SZ
/ 4;
201 /* tsk_inc(): increment counter for sent or received data
202 * - If block based flow control is not supported by peer we
203 * fall back to message based ditto, incrementing the counter
205 static u16
tsk_inc(struct tipc_sock
*tsk
, int msglen
)
207 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
208 return ((msglen
/ FLOWCTL_BLK_SZ
) + 1);
213 * tsk_advance_rx_queue - discard first buffer in socket receive queue
215 * Caller must hold socket lock
217 static void tsk_advance_rx_queue(struct sock
*sk
)
219 kfree_skb(__skb_dequeue(&sk
->sk_receive_queue
));
222 /* tipc_sk_respond() : send response message back to sender
224 static void tipc_sk_respond(struct sock
*sk
, struct sk_buff
*skb
, int err
)
228 u32 onode
= tipc_own_addr(sock_net(sk
));
230 if (!tipc_msg_reverse(onode
, &skb
, err
))
233 dnode
= msg_destnode(buf_msg(skb
));
234 selector
= msg_origport(buf_msg(skb
));
235 tipc_node_xmit_skb(sock_net(sk
), skb
, dnode
, selector
);
239 * tsk_rej_rx_queue - reject all buffers in socket receive queue
241 * Caller must hold socket lock
243 static void tsk_rej_rx_queue(struct sock
*sk
)
247 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)))
248 tipc_sk_respond(sk
, skb
, TIPC_ERR_NO_PORT
);
251 static bool tipc_sk_connected(struct sock
*sk
)
253 return sk
->sk_state
== TIPC_ESTABLISHED
;
256 /* tipc_sk_type_connectionless - check if the socket is datagram socket
259 * Returns true if connection less, false otherwise
261 static bool tipc_sk_type_connectionless(struct sock
*sk
)
263 return sk
->sk_type
== SOCK_RDM
|| sk
->sk_type
== SOCK_DGRAM
;
266 /* tsk_peer_msg - verify if message was sent by connected port's peer
268 * Handles cases where the node's network address has changed from
269 * the default of <0.0.0> to its configured setting.
271 static bool tsk_peer_msg(struct tipc_sock
*tsk
, struct tipc_msg
*msg
)
273 struct sock
*sk
= &tsk
->sk
;
274 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
275 u32 peer_port
= tsk_peer_port(tsk
);
279 if (unlikely(!tipc_sk_connected(sk
)))
282 if (unlikely(msg_origport(msg
) != peer_port
))
285 orig_node
= msg_orignode(msg
);
286 peer_node
= tsk_peer_node(tsk
);
288 if (likely(orig_node
== peer_node
))
291 if (!orig_node
&& (peer_node
== tn
->own_addr
))
294 if (!peer_node
&& (orig_node
== tn
->own_addr
))
300 /* tipc_set_sk_state - set the sk_state of the socket
303 * Caller must hold socket lock
305 * Returns 0 on success, errno otherwise
307 static int tipc_set_sk_state(struct sock
*sk
, int state
)
309 int oldsk_state
= sk
->sk_state
;
317 case TIPC_CONNECTING
:
318 if (oldsk_state
== TIPC_OPEN
)
321 case TIPC_ESTABLISHED
:
322 if (oldsk_state
== TIPC_CONNECTING
||
323 oldsk_state
== TIPC_OPEN
)
326 case TIPC_DISCONNECTING
:
327 if (oldsk_state
== TIPC_CONNECTING
||
328 oldsk_state
== TIPC_ESTABLISHED
)
334 sk
->sk_state
= state
;
339 static int tipc_sk_sock_err(struct socket
*sock
, long *timeout
)
341 struct sock
*sk
= sock
->sk
;
342 int err
= sock_error(sk
);
343 int typ
= sock
->type
;
347 if (typ
== SOCK_STREAM
|| typ
== SOCK_SEQPACKET
) {
348 if (sk
->sk_state
== TIPC_DISCONNECTING
)
350 else if (!tipc_sk_connected(sk
))
355 if (signal_pending(current
))
356 return sock_intr_errno(*timeout
);
361 #define tipc_wait_for_cond(sock_, timeout_, condition_) \
366 while (!(condition_) && !done_) { \
367 struct sock *sk_ = sock->sk; \
368 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
370 rc_ = tipc_sk_sock_err(sock_, timeout_); \
373 prepare_to_wait(sk_sleep(sk_), &wait_, \
374 TASK_INTERRUPTIBLE); \
375 done_ = sk_wait_event(sk_, timeout_, \
376 (condition_), &wait_); \
377 remove_wait_queue(sk_sleep(sk_), &wait_); \
383 * tipc_sk_create - create a TIPC socket
384 * @net: network namespace (must be default network)
385 * @sock: pre-allocated socket structure
386 * @protocol: protocol indicator (must be 0)
387 * @kern: caused by kernel or by userspace?
389 * This routine creates additional data structures used by the TIPC socket,
390 * initializes them, and links them together.
392 * Returns 0 on success, errno otherwise
394 static int tipc_sk_create(struct net
*net
, struct socket
*sock
,
395 int protocol
, int kern
)
398 const struct proto_ops
*ops
;
400 struct tipc_sock
*tsk
;
401 struct tipc_msg
*msg
;
403 /* Validate arguments */
404 if (unlikely(protocol
!= 0))
405 return -EPROTONOSUPPORT
;
407 switch (sock
->type
) {
422 /* Allocate socket's protocol area */
423 sk
= sk_alloc(net
, AF_TIPC
, GFP_KERNEL
, &tipc_proto
, kern
);
428 tsk
->max_pkt
= MAX_PKT_DEFAULT
;
429 INIT_LIST_HEAD(&tsk
->publications
);
430 INIT_LIST_HEAD(&tsk
->cong_links
);
432 tn
= net_generic(sock_net(sk
), tipc_net_id
);
434 /* Finish initializing socket data structures */
436 sock_init_data(sock
, sk
);
437 tipc_set_sk_state(sk
, TIPC_OPEN
);
438 if (tipc_sk_insert(tsk
)) {
439 pr_warn("Socket create failed; port number exhausted\n");
443 /* Ensure tsk is visible before we read own_addr. */
446 tipc_msg_init(tn
->own_addr
, msg
, TIPC_LOW_IMPORTANCE
, TIPC_NAMED_MSG
,
449 msg_set_origport(msg
, tsk
->portid
);
450 setup_timer(&sk
->sk_timer
, tipc_sk_timeout
, (unsigned long)tsk
);
452 sk
->sk_backlog_rcv
= tipc_backlog_rcv
;
453 sk
->sk_rcvbuf
= sysctl_tipc_rmem
[1];
454 sk
->sk_data_ready
= tipc_data_ready
;
455 sk
->sk_write_space
= tipc_write_space
;
456 sk
->sk_destruct
= tipc_sock_destruct
;
457 tsk
->conn_timeout
= CONN_TIMEOUT_DEFAULT
;
458 atomic_set(&tsk
->dupl_rcvcnt
, 0);
460 /* Start out with safe limits until we receive an advertised window */
461 tsk
->snd_win
= tsk_adv_blocks(RCVBUF_MIN
);
462 tsk
->rcv_win
= tsk
->snd_win
;
464 if (tipc_sk_type_connectionless(sk
)) {
465 tsk_set_unreturnable(tsk
, true);
466 if (sock
->type
== SOCK_DGRAM
)
467 tsk_set_unreliable(tsk
, true);
473 static void tipc_sk_callback(struct rcu_head
*head
)
475 struct tipc_sock
*tsk
= container_of(head
, struct tipc_sock
, rcu
);
480 /* Caller should hold socket lock for the socket. */
481 static void __tipc_shutdown(struct socket
*sock
, int error
)
483 struct sock
*sk
= sock
->sk
;
484 struct tipc_sock
*tsk
= tipc_sk(sk
);
485 struct net
*net
= sock_net(sk
);
486 long timeout
= CONN_TIMEOUT_DEFAULT
;
487 u32 dnode
= tsk_peer_node(tsk
);
490 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
491 tipc_wait_for_cond(sock
, &timeout
, (!tsk
->cong_link_cnt
&&
492 !tsk_conn_cong(tsk
)));
494 /* Reject all unreceived messages, except on an active connection
495 * (which disconnects locally & sends a 'FIN+' to peer).
497 while ((skb
= __skb_dequeue(&sk
->sk_receive_queue
)) != NULL
) {
498 if (TIPC_SKB_CB(skb
)->bytes_read
) {
502 if (!tipc_sk_type_connectionless(sk
) &&
503 sk
->sk_state
!= TIPC_DISCONNECTING
) {
504 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
505 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
507 tipc_sk_respond(sk
, skb
, error
);
510 if (tipc_sk_type_connectionless(sk
))
513 if (sk
->sk_state
!= TIPC_DISCONNECTING
) {
514 skb
= tipc_msg_create(TIPC_CRITICAL_IMPORTANCE
,
515 TIPC_CONN_MSG
, SHORT_H_SIZE
, 0, dnode
,
516 tsk_own_node(tsk
), tsk_peer_port(tsk
),
519 tipc_node_xmit_skb(net
, skb
, dnode
, tsk
->portid
);
520 tipc_node_remove_conn(net
, dnode
, tsk
->portid
);
521 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
526 * tipc_release - destroy a TIPC socket
527 * @sock: socket to destroy
529 * This routine cleans up any messages that are still queued on the socket.
530 * For DGRAM and RDM socket types, all queued messages are rejected.
531 * For SEQPACKET and STREAM socket types, the first message is rejected
532 * and any others are discarded. (If the first message on a STREAM socket
533 * is partially-read, it is discarded and the next one is rejected instead.)
535 * NOTE: Rejected messages are not necessarily returned to the sender! They
536 * are returned or discarded according to the "destination droppable" setting
537 * specified for the message by the sender.
539 * Returns 0 on success, errno otherwise
541 static int tipc_release(struct socket
*sock
)
543 struct sock
*sk
= sock
->sk
;
544 struct tipc_sock
*tsk
;
547 * Exit if socket isn't fully initialized (occurs when a failed accept()
548 * releases a pre-allocated child socket that was never used)
556 __tipc_shutdown(sock
, TIPC_ERR_NO_PORT
);
557 sk
->sk_shutdown
= SHUTDOWN_MASK
;
558 tipc_sk_withdraw(tsk
, 0, NULL
);
559 sk_stop_timer(sk
, &sk
->sk_timer
);
562 /* Reject any messages that accumulated in backlog queue */
564 u32_list_purge(&tsk
->cong_links
);
565 tsk
->cong_link_cnt
= 0;
566 call_rcu(&tsk
->rcu
, tipc_sk_callback
);
573 * tipc_bind - associate or disassocate TIPC name(s) with a socket
574 * @sock: socket structure
575 * @uaddr: socket address describing name(s) and desired operation
576 * @uaddr_len: size of socket address data structure
578 * Name and name sequence binding is indicated using a positive scope value;
579 * a negative scope value unbinds the specified name. Specifying no name
580 * (i.e. a socket address length of 0) unbinds all names from the socket.
582 * Returns 0 on success, errno otherwise
584 * NOTE: This routine doesn't need to take the socket lock since it doesn't
585 * access any non-constant socket information.
587 static int tipc_bind(struct socket
*sock
, struct sockaddr
*uaddr
,
590 struct sock
*sk
= sock
->sk
;
591 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
592 struct tipc_sock
*tsk
= tipc_sk(sk
);
596 if (unlikely(!uaddr_len
)) {
597 res
= tipc_sk_withdraw(tsk
, 0, NULL
);
601 if (uaddr_len
< sizeof(struct sockaddr_tipc
)) {
605 if (addr
->family
!= AF_TIPC
) {
610 if (addr
->addrtype
== TIPC_ADDR_NAME
)
611 addr
->addr
.nameseq
.upper
= addr
->addr
.nameseq
.lower
;
612 else if (addr
->addrtype
!= TIPC_ADDR_NAMESEQ
) {
617 if ((addr
->addr
.nameseq
.type
< TIPC_RESERVED_TYPES
) &&
618 (addr
->addr
.nameseq
.type
!= TIPC_TOP_SRV
) &&
619 (addr
->addr
.nameseq
.type
!= TIPC_CFG_SRV
)) {
624 res
= (addr
->scope
> 0) ?
625 tipc_sk_publish(tsk
, addr
->scope
, &addr
->addr
.nameseq
) :
626 tipc_sk_withdraw(tsk
, -addr
->scope
, &addr
->addr
.nameseq
);
633 * tipc_getname - get port ID of socket or peer socket
634 * @sock: socket structure
635 * @uaddr: area for returned socket address
636 * @uaddr_len: area for returned length of socket address
637 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
639 * Returns 0 on success, errno otherwise
641 * NOTE: This routine doesn't need to take the socket lock since it only
642 * accesses socket information that is unchanging (or which changes in
643 * a completely predictable manner).
645 static int tipc_getname(struct socket
*sock
, struct sockaddr
*uaddr
,
646 int *uaddr_len
, int peer
)
648 struct sockaddr_tipc
*addr
= (struct sockaddr_tipc
*)uaddr
;
649 struct sock
*sk
= sock
->sk
;
650 struct tipc_sock
*tsk
= tipc_sk(sk
);
651 struct tipc_net
*tn
= net_generic(sock_net(sock
->sk
), tipc_net_id
);
653 memset(addr
, 0, sizeof(*addr
));
655 if ((!tipc_sk_connected(sk
)) &&
656 ((peer
!= 2) || (sk
->sk_state
!= TIPC_DISCONNECTING
)))
658 addr
->addr
.id
.ref
= tsk_peer_port(tsk
);
659 addr
->addr
.id
.node
= tsk_peer_node(tsk
);
661 addr
->addr
.id
.ref
= tsk
->portid
;
662 addr
->addr
.id
.node
= tn
->own_addr
;
665 *uaddr_len
= sizeof(*addr
);
666 addr
->addrtype
= TIPC_ADDR_ID
;
667 addr
->family
= AF_TIPC
;
669 addr
->addr
.name
.domain
= 0;
675 * tipc_poll - read and possibly block on pollmask
676 * @file: file structure associated with the socket
677 * @sock: socket for which to calculate the poll bits
680 * Returns pollmask value
683 * It appears that the usual socket locking mechanisms are not useful here
684 * since the pollmask info is potentially out-of-date the moment this routine
685 * exits. TCP and other protocols seem to rely on higher level poll routines
686 * to handle any preventable race conditions, so TIPC will do the same ...
688 * IMPORTANT: The fact that a read or write operation is indicated does NOT
689 * imply that the operation will succeed, merely that it should be performed
690 * and will not block.
692 static unsigned int tipc_poll(struct file
*file
, struct socket
*sock
,
695 struct sock
*sk
= sock
->sk
;
696 struct tipc_sock
*tsk
= tipc_sk(sk
);
699 sock_poll_wait(file
, sk_sleep(sk
), wait
);
701 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
702 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
703 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
706 switch (sk
->sk_state
) {
707 case TIPC_ESTABLISHED
:
708 if (!tsk
->cong_link_cnt
&& !tsk_conn_cong(tsk
))
712 case TIPC_CONNECTING
:
713 if (!skb_queue_empty(&sk
->sk_receive_queue
))
714 mask
|= (POLLIN
| POLLRDNORM
);
717 if (!tsk
->cong_link_cnt
)
719 if (tipc_sk_type_connectionless(sk
) &&
720 (!skb_queue_empty(&sk
->sk_receive_queue
)))
721 mask
|= (POLLIN
| POLLRDNORM
);
723 case TIPC_DISCONNECTING
:
724 mask
= (POLLIN
| POLLRDNORM
| POLLHUP
);
732 * tipc_sendmcast - send multicast message
733 * @sock: socket structure
734 * @seq: destination address
735 * @msg: message to send
736 * @dlen: length of data to send
737 * @timeout: timeout to wait for wakeup
739 * Called from function tipc_sendmsg(), which has done all sanity checks
740 * Returns the number of bytes sent on success, or errno
742 static int tipc_sendmcast(struct socket
*sock
, struct tipc_name_seq
*seq
,
743 struct msghdr
*msg
, size_t dlen
, long timeout
)
745 struct sock
*sk
= sock
->sk
;
746 struct tipc_sock
*tsk
= tipc_sk(sk
);
747 struct tipc_msg
*hdr
= &tsk
->phdr
;
748 struct net
*net
= sock_net(sk
);
749 int mtu
= tipc_bcast_get_mtu(net
);
750 struct tipc_mc_method
*method
= &tsk
->mc_method
;
751 u32 domain
= addr_domain(net
, TIPC_CLUSTER_SCOPE
);
752 struct sk_buff_head pkts
;
753 struct tipc_nlist dsts
;
756 /* Block or return if any destination link is congested */
757 rc
= tipc_wait_for_cond(sock
, &timeout
, !tsk
->cong_link_cnt
);
761 /* Lookup destination nodes */
762 tipc_nlist_init(&dsts
, tipc_own_addr(net
));
763 tipc_nametbl_lookup_dst_nodes(net
, seq
->type
, seq
->lower
,
764 seq
->upper
, domain
, &dsts
);
765 if (!dsts
.local
&& !dsts
.remote
)
766 return -EHOSTUNREACH
;
768 /* Build message header */
769 msg_set_type(hdr
, TIPC_MCAST_MSG
);
770 msg_set_hdr_sz(hdr
, MCAST_H_SIZE
);
771 msg_set_lookup_scope(hdr
, TIPC_CLUSTER_SCOPE
);
772 msg_set_destport(hdr
, 0);
773 msg_set_destnode(hdr
, 0);
774 msg_set_nametype(hdr
, seq
->type
);
775 msg_set_namelower(hdr
, seq
->lower
);
776 msg_set_nameupper(hdr
, seq
->upper
);
778 /* Build message as chain of buffers */
779 skb_queue_head_init(&pkts
);
780 rc
= tipc_msg_build(hdr
, msg
, 0, dlen
, mtu
, &pkts
);
782 /* Send message if build was successful */
783 if (unlikely(rc
== dlen
))
784 rc
= tipc_mcast_xmit(net
, &pkts
, method
, &dsts
,
785 &tsk
->cong_link_cnt
);
787 tipc_nlist_purge(&dsts
);
789 return rc
? rc
: dlen
;
793 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
794 * @arrvq: queue with arriving messages, to be cloned after destination lookup
795 * @inputq: queue with cloned messages, delivered to socket after dest lookup
797 * Multi-threaded: parallel calls with reference to same queues may occur
799 void tipc_sk_mcast_rcv(struct net
*net
, struct sk_buff_head
*arrvq
,
800 struct sk_buff_head
*inputq
)
802 struct tipc_msg
*msg
;
803 struct list_head dports
;
805 u32 scope
= TIPC_CLUSTER_SCOPE
;
806 struct sk_buff_head tmpq
;
808 struct sk_buff
*skb
, *_skb
;
810 __skb_queue_head_init(&tmpq
);
811 INIT_LIST_HEAD(&dports
);
813 skb
= tipc_skb_peek(arrvq
, &inputq
->lock
);
814 for (; skb
; skb
= tipc_skb_peek(arrvq
, &inputq
->lock
)) {
816 hsz
= skb_headroom(skb
) + msg_hdr_sz(msg
);
818 if (in_own_node(net
, msg_orignode(msg
)))
819 scope
= TIPC_NODE_SCOPE
;
821 /* Create destination port list and message clones: */
822 tipc_nametbl_mc_translate(net
,
823 msg_nametype(msg
), msg_namelower(msg
),
824 msg_nameupper(msg
), scope
, &dports
);
825 portid
= u32_pop(&dports
);
826 for (; portid
; portid
= u32_pop(&dports
)) {
827 _skb
= __pskb_copy(skb
, hsz
, GFP_ATOMIC
);
829 msg_set_destport(buf_msg(_skb
), portid
);
830 __skb_queue_tail(&tmpq
, _skb
);
833 pr_warn("Failed to clone mcast rcv buffer\n");
835 /* Append to inputq if not already done by other thread */
836 spin_lock_bh(&inputq
->lock
);
837 if (skb_peek(arrvq
) == skb
) {
838 skb_queue_splice_tail_init(&tmpq
, inputq
);
839 kfree_skb(__skb_dequeue(arrvq
));
841 spin_unlock_bh(&inputq
->lock
);
842 __skb_queue_purge(&tmpq
);
845 tipc_sk_rcv(net
, inputq
);
849 * tipc_sk_proto_rcv - receive a connection mng protocol message
850 * @tsk: receiving socket
851 * @skb: pointer to message buffer.
853 static void tipc_sk_proto_rcv(struct tipc_sock
*tsk
, struct sk_buff
*skb
,
854 struct sk_buff_head
*xmitq
)
856 struct sock
*sk
= &tsk
->sk
;
857 u32 onode
= tsk_own_node(tsk
);
858 struct tipc_msg
*hdr
= buf_msg(skb
);
859 int mtyp
= msg_type(hdr
);
862 /* Ignore if connection cannot be validated: */
863 if (!tsk_peer_msg(tsk
, hdr
))
866 tsk
->probe_unacked
= false;
868 if (mtyp
== CONN_PROBE
) {
869 msg_set_type(hdr
, CONN_PROBE_REPLY
);
870 if (tipc_msg_reverse(onode
, &skb
, TIPC_OK
))
871 __skb_queue_tail(xmitq
, skb
);
873 } else if (mtyp
== CONN_ACK
) {
874 conn_cong
= tsk_conn_cong(tsk
);
875 tsk
->snt_unacked
-= msg_conn_ack(hdr
);
876 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
877 tsk
->snd_win
= msg_adv_win(hdr
);
879 sk
->sk_write_space(sk
);
880 } else if (mtyp
!= CONN_PROBE_REPLY
) {
881 pr_warn("Received unknown CONN_PROTO msg\n");
888 * tipc_sendmsg - send message in connectionless manner
889 * @sock: socket structure
890 * @m: message to send
891 * @dsz: amount of user data to be sent
893 * Message must have an destination specified explicitly.
894 * Used for SOCK_RDM and SOCK_DGRAM messages,
895 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
896 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
898 * Returns the number of bytes sent on success, or errno otherwise
900 static int tipc_sendmsg(struct socket
*sock
,
901 struct msghdr
*m
, size_t dsz
)
903 struct sock
*sk
= sock
->sk
;
907 ret
= __tipc_sendmsg(sock
, m
, dsz
);
913 static int __tipc_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
915 struct sock
*sk
= sock
->sk
;
916 struct net
*net
= sock_net(sk
);
917 struct tipc_sock
*tsk
= tipc_sk(sk
);
918 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
919 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
920 struct list_head
*clinks
= &tsk
->cong_links
;
921 bool syn
= !tipc_sk_type_connectionless(sk
);
922 struct tipc_msg
*hdr
= &tsk
->phdr
;
923 struct tipc_name_seq
*seq
;
924 struct sk_buff_head pkts
;
925 u32 type
, inst
, domain
;
929 if (unlikely(dlen
> TIPC_MAX_USER_MSG_SIZE
))
932 if (unlikely(!dest
)) {
934 if (!syn
|| dest
->family
!= AF_TIPC
)
935 return -EDESTADDRREQ
;
938 if (unlikely(m
->msg_namelen
< sizeof(*dest
)))
941 if (unlikely(dest
->family
!= AF_TIPC
))
945 if (sk
->sk_state
== TIPC_LISTEN
)
947 if (sk
->sk_state
!= TIPC_OPEN
)
951 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
952 tsk
->conn_type
= dest
->addr
.name
.name
.type
;
953 tsk
->conn_instance
= dest
->addr
.name
.name
.instance
;
957 seq
= &dest
->addr
.nameseq
;
958 if (dest
->addrtype
== TIPC_ADDR_MCAST
)
959 return tipc_sendmcast(sock
, seq
, m
, dlen
, timeout
);
961 if (dest
->addrtype
== TIPC_ADDR_NAME
) {
962 type
= dest
->addr
.name
.name
.type
;
963 inst
= dest
->addr
.name
.name
.instance
;
964 domain
= dest
->addr
.name
.domain
;
966 msg_set_type(hdr
, TIPC_NAMED_MSG
);
967 msg_set_hdr_sz(hdr
, NAMED_H_SIZE
);
968 msg_set_nametype(hdr
, type
);
969 msg_set_nameinst(hdr
, inst
);
970 msg_set_lookup_scope(hdr
, tipc_addr_scope(domain
));
971 dport
= tipc_nametbl_translate(net
, type
, inst
, &dnode
);
972 msg_set_destnode(hdr
, dnode
);
973 msg_set_destport(hdr
, dport
);
974 if (unlikely(!dport
&& !dnode
))
975 return -EHOSTUNREACH
;
977 } else if (dest
->addrtype
== TIPC_ADDR_ID
) {
978 dnode
= dest
->addr
.id
.node
;
979 msg_set_type(hdr
, TIPC_DIRECT_MSG
);
980 msg_set_lookup_scope(hdr
, 0);
981 msg_set_destnode(hdr
, dnode
);
982 msg_set_destport(hdr
, dest
->addr
.id
.ref
);
983 msg_set_hdr_sz(hdr
, BASIC_H_SIZE
);
986 /* Block or return if destination link is congested */
987 rc
= tipc_wait_for_cond(sock
, &timeout
, !u32_find(clinks
, dnode
));
991 skb_queue_head_init(&pkts
);
992 mtu
= tipc_node_get_mtu(net
, dnode
, tsk
->portid
);
993 rc
= tipc_msg_build(hdr
, m
, 0, dlen
, mtu
, &pkts
);
994 if (unlikely(rc
!= dlen
))
997 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
998 if (unlikely(rc
== -ELINKCONG
)) {
999 u32_push(clinks
, dnode
);
1000 tsk
->cong_link_cnt
++;
1004 if (unlikely(syn
&& !rc
))
1005 tipc_set_sk_state(sk
, TIPC_CONNECTING
);
1007 return rc
? rc
: dlen
;
1011 * tipc_sendstream - send stream-oriented data
1012 * @sock: socket structure
1014 * @dsz: total length of data to be transmitted
1016 * Used for SOCK_STREAM data.
1018 * Returns the number of bytes sent on success (or partial success),
1019 * or errno if no data sent
1021 static int tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1023 struct sock
*sk
= sock
->sk
;
1027 ret
= __tipc_sendstream(sock
, m
, dsz
);
1033 static int __tipc_sendstream(struct socket
*sock
, struct msghdr
*m
, size_t dlen
)
1035 struct sock
*sk
= sock
->sk
;
1036 DECLARE_SOCKADDR(struct sockaddr_tipc
*, dest
, m
->msg_name
);
1037 long timeout
= sock_sndtimeo(sk
, m
->msg_flags
& MSG_DONTWAIT
);
1038 struct tipc_sock
*tsk
= tipc_sk(sk
);
1039 struct tipc_msg
*hdr
= &tsk
->phdr
;
1040 struct net
*net
= sock_net(sk
);
1041 struct sk_buff_head pkts
;
1042 u32 dnode
= tsk_peer_node(tsk
);
1046 skb_queue_head_init(&pkts
);
1048 if (unlikely(dlen
> INT_MAX
))
1051 /* Handle implicit connection setup */
1052 if (unlikely(dest
)) {
1053 rc
= __tipc_sendmsg(sock
, m
, dlen
);
1054 if (dlen
&& (dlen
== rc
))
1055 tsk
->snt_unacked
= tsk_inc(tsk
, dlen
+ msg_hdr_sz(hdr
));
1060 rc
= tipc_wait_for_cond(sock
, &timeout
,
1061 (!tsk
->cong_link_cnt
&&
1062 !tsk_conn_cong(tsk
) &&
1063 tipc_sk_connected(sk
)));
1067 send
= min_t(size_t, dlen
- sent
, TIPC_MAX_USER_MSG_SIZE
);
1068 rc
= tipc_msg_build(hdr
, m
, sent
, send
, tsk
->max_pkt
, &pkts
);
1069 if (unlikely(rc
!= send
))
1072 rc
= tipc_node_xmit(net
, &pkts
, dnode
, tsk
->portid
);
1073 if (unlikely(rc
== -ELINKCONG
)) {
1074 tsk
->cong_link_cnt
= 1;
1078 tsk
->snt_unacked
+= tsk_inc(tsk
, send
+ MIN_H_SIZE
);
1081 } while (sent
< dlen
&& !rc
);
1083 return rc
? rc
: sent
;
1087 * tipc_send_packet - send a connection-oriented message
1088 * @sock: socket structure
1089 * @m: message to send
1090 * @dsz: length of data to be transmitted
1092 * Used for SOCK_SEQPACKET messages.
1094 * Returns the number of bytes sent on success, or errno otherwise
1096 static int tipc_send_packet(struct socket
*sock
, struct msghdr
*m
, size_t dsz
)
1098 if (dsz
> TIPC_MAX_USER_MSG_SIZE
)
1101 return tipc_sendstream(sock
, m
, dsz
);
1104 /* tipc_sk_finish_conn - complete the setup of a connection
1106 static void tipc_sk_finish_conn(struct tipc_sock
*tsk
, u32 peer_port
,
1109 struct sock
*sk
= &tsk
->sk
;
1110 struct net
*net
= sock_net(sk
);
1111 struct tipc_msg
*msg
= &tsk
->phdr
;
1113 msg_set_destnode(msg
, peer_node
);
1114 msg_set_destport(msg
, peer_port
);
1115 msg_set_type(msg
, TIPC_CONN_MSG
);
1116 msg_set_lookup_scope(msg
, 0);
1117 msg_set_hdr_sz(msg
, SHORT_H_SIZE
);
1119 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTERVAL
);
1120 tipc_set_sk_state(sk
, TIPC_ESTABLISHED
);
1121 tipc_node_add_conn(net
, peer_node
, tsk
->portid
, peer_port
);
1122 tsk
->max_pkt
= tipc_node_get_mtu(net
, peer_node
, tsk
->portid
);
1123 tsk
->peer_caps
= tipc_node_get_capabilities(net
, peer_node
);
1124 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
)
1127 /* Fall back to message based flow control */
1128 tsk
->rcv_win
= FLOWCTL_MSG_WIN
;
1129 tsk
->snd_win
= FLOWCTL_MSG_WIN
;
1133 * set_orig_addr - capture sender's address for received message
1134 * @m: descriptor for message info
1135 * @msg: received message header
1137 * Note: Address is not captured if not requested by receiver.
1139 static void set_orig_addr(struct msghdr
*m
, struct tipc_msg
*msg
)
1141 DECLARE_SOCKADDR(struct sockaddr_tipc
*, addr
, m
->msg_name
);
1144 addr
->family
= AF_TIPC
;
1145 addr
->addrtype
= TIPC_ADDR_ID
;
1146 memset(&addr
->addr
, 0, sizeof(addr
->addr
));
1147 addr
->addr
.id
.ref
= msg_origport(msg
);
1148 addr
->addr
.id
.node
= msg_orignode(msg
);
1149 addr
->addr
.name
.domain
= 0; /* could leave uninitialized */
1150 addr
->scope
= 0; /* could leave uninitialized */
1151 m
->msg_namelen
= sizeof(struct sockaddr_tipc
);
1156 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1157 * @m: descriptor for message info
1158 * @msg: received message header
1159 * @tsk: TIPC port associated with message
1161 * Note: Ancillary data is not captured if not requested by receiver.
1163 * Returns 0 if successful, otherwise errno
1165 static int tipc_sk_anc_data_recv(struct msghdr
*m
, struct tipc_msg
*msg
,
1166 struct tipc_sock
*tsk
)
1174 if (likely(m
->msg_controllen
== 0))
1177 /* Optionally capture errored message object(s) */
1178 err
= msg
? msg_errcode(msg
) : 0;
1179 if (unlikely(err
)) {
1181 anc_data
[1] = msg_data_sz(msg
);
1182 res
= put_cmsg(m
, SOL_TIPC
, TIPC_ERRINFO
, 8, anc_data
);
1186 res
= put_cmsg(m
, SOL_TIPC
, TIPC_RETDATA
, anc_data
[1],
1193 /* Optionally capture message destination object */
1194 dest_type
= msg
? msg_type(msg
) : TIPC_DIRECT_MSG
;
1195 switch (dest_type
) {
1196 case TIPC_NAMED_MSG
:
1198 anc_data
[0] = msg_nametype(msg
);
1199 anc_data
[1] = msg_namelower(msg
);
1200 anc_data
[2] = msg_namelower(msg
);
1202 case TIPC_MCAST_MSG
:
1204 anc_data
[0] = msg_nametype(msg
);
1205 anc_data
[1] = msg_namelower(msg
);
1206 anc_data
[2] = msg_nameupper(msg
);
1209 has_name
= (tsk
->conn_type
!= 0);
1210 anc_data
[0] = tsk
->conn_type
;
1211 anc_data
[1] = tsk
->conn_instance
;
1212 anc_data
[2] = tsk
->conn_instance
;
1218 res
= put_cmsg(m
, SOL_TIPC
, TIPC_DESTNAME
, 12, anc_data
);
1226 static void tipc_sk_send_ack(struct tipc_sock
*tsk
)
1228 struct sock
*sk
= &tsk
->sk
;
1229 struct net
*net
= sock_net(sk
);
1230 struct sk_buff
*skb
= NULL
;
1231 struct tipc_msg
*msg
;
1232 u32 peer_port
= tsk_peer_port(tsk
);
1233 u32 dnode
= tsk_peer_node(tsk
);
1235 if (!tipc_sk_connected(sk
))
1237 skb
= tipc_msg_create(CONN_MANAGER
, CONN_ACK
, INT_H_SIZE
, 0,
1238 dnode
, tsk_own_node(tsk
), peer_port
,
1239 tsk
->portid
, TIPC_OK
);
1243 msg_set_conn_ack(msg
, tsk
->rcv_unacked
);
1244 tsk
->rcv_unacked
= 0;
1246 /* Adjust to and advertize the correct window limit */
1247 if (tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
) {
1248 tsk
->rcv_win
= tsk_adv_blocks(tsk
->sk
.sk_rcvbuf
);
1249 msg_set_adv_win(msg
, tsk
->rcv_win
);
1251 tipc_node_xmit_skb(net
, skb
, dnode
, msg_link_selector(msg
));
1254 static int tipc_wait_for_rcvmsg(struct socket
*sock
, long *timeop
)
1256 struct sock
*sk
= sock
->sk
;
1258 long timeo
= *timeop
;
1262 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1263 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
1264 if (sk
->sk_shutdown
& RCV_SHUTDOWN
) {
1269 timeo
= schedule_timeout(timeo
);
1273 if (!skb_queue_empty(&sk
->sk_receive_queue
))
1278 err
= sock_intr_errno(timeo
);
1279 if (signal_pending(current
))
1282 finish_wait(sk_sleep(sk
), &wait
);
1288 * tipc_recvmsg - receive packet-oriented message
1289 * @m: descriptor for message info
1290 * @buf_len: total size of user buffer area
1291 * @flags: receive flags
1293 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1294 * If the complete message doesn't fit in user area, truncate it.
1296 * Returns size of returned message data, errno otherwise
1298 static int tipc_recvmsg(struct socket
*sock
, struct msghdr
*m
, size_t buf_len
,
1301 struct sock
*sk
= sock
->sk
;
1302 struct tipc_sock
*tsk
= tipc_sk(sk
);
1303 struct sk_buff
*buf
;
1304 struct tipc_msg
*msg
;
1305 bool is_connectionless
= tipc_sk_type_connectionless(sk
);
1311 /* Catch invalid receive requests */
1312 if (unlikely(!buf_len
))
1317 if (!is_connectionless
&& unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1322 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1325 /* Look for a message in receive queue; wait if necessary */
1326 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1330 /* Look at first message in receive queue */
1331 buf
= skb_peek(&sk
->sk_receive_queue
);
1333 sz
= msg_data_sz(msg
);
1334 hlen
= msg_hdr_sz(msg
);
1335 err
= msg_errcode(msg
);
1337 /* Discard an empty non-errored message & try again */
1338 if ((!sz
) && (!err
)) {
1339 tsk_advance_rx_queue(sk
);
1343 /* Capture sender's address (optional) */
1344 set_orig_addr(m
, msg
);
1346 /* Capture ancillary data (optional) */
1347 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1351 /* Capture message data (if valid) & compute return value (always) */
1353 if (unlikely(buf_len
< sz
)) {
1355 m
->msg_flags
|= MSG_TRUNC
;
1357 res
= skb_copy_datagram_msg(buf
, hlen
, m
, sz
);
1362 if (is_connectionless
|| err
== TIPC_CONN_SHUTDOWN
||
1369 if (unlikely(flags
& MSG_PEEK
))
1372 if (likely(!is_connectionless
)) {
1373 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ sz
);
1374 if (unlikely(tsk
->rcv_unacked
>= (tsk
->rcv_win
/ 4)))
1375 tipc_sk_send_ack(tsk
);
1377 tsk_advance_rx_queue(sk
);
1384 * tipc_recv_stream - receive stream-oriented data
1385 * @m: descriptor for message info
1386 * @buf_len: total size of user buffer area
1387 * @flags: receive flags
1389 * Used for SOCK_STREAM messages only. If not enough data is available
1390 * will optionally wait for more; never truncates data.
1392 * Returns size of returned message data, errno otherwise
1394 static int tipc_recv_stream(struct socket
*sock
, struct msghdr
*m
,
1395 size_t buf_len
, int flags
)
1397 struct sock
*sk
= sock
->sk
;
1398 struct tipc_sock
*tsk
= tipc_sk(sk
);
1399 struct sk_buff
*buf
;
1400 struct tipc_msg
*msg
;
1408 /* Catch invalid receive attempts */
1409 if (unlikely(!buf_len
))
1414 if (unlikely(sk
->sk_state
== TIPC_OPEN
)) {
1419 target
= sock_rcvlowat(sk
, flags
& MSG_WAITALL
, buf_len
);
1420 timeo
= sock_rcvtimeo(sk
, flags
& MSG_DONTWAIT
);
1423 /* Look for a message in receive queue; wait if necessary */
1424 res
= tipc_wait_for_rcvmsg(sock
, &timeo
);
1428 /* Look at first message in receive queue */
1429 buf
= skb_peek(&sk
->sk_receive_queue
);
1431 sz
= msg_data_sz(msg
);
1432 hlen
= msg_hdr_sz(msg
);
1433 err
= msg_errcode(msg
);
1435 /* Discard an empty non-errored message & try again */
1436 if ((!sz
) && (!err
)) {
1437 tsk_advance_rx_queue(sk
);
1441 /* Optionally capture sender's address & ancillary data of first msg */
1442 if (sz_copied
== 0) {
1443 set_orig_addr(m
, msg
);
1444 res
= tipc_sk_anc_data_recv(m
, msg
, tsk
);
1449 /* Capture message data (if valid) & compute return value (always) */
1451 u32 offset
= TIPC_SKB_CB(buf
)->bytes_read
;
1456 needed
= (buf_len
- sz_copied
);
1457 sz_to_copy
= min(sz
, needed
);
1459 res
= skb_copy_datagram_msg(buf
, hlen
+ offset
, m
, sz_to_copy
);
1463 sz_copied
+= sz_to_copy
;
1465 if (sz_to_copy
< sz
) {
1466 if (!(flags
& MSG_PEEK
))
1467 TIPC_SKB_CB(buf
)->bytes_read
=
1468 offset
+ sz_to_copy
;
1473 goto exit
; /* can't add error msg to valid data */
1475 if ((err
== TIPC_CONN_SHUTDOWN
) || m
->msg_control
)
1481 if (unlikely(flags
& MSG_PEEK
))
1484 tsk
->rcv_unacked
+= tsk_inc(tsk
, hlen
+ sz
);
1485 if (unlikely(tsk
->rcv_unacked
>= (tsk
->rcv_win
/ 4)))
1486 tipc_sk_send_ack(tsk
);
1487 tsk_advance_rx_queue(sk
);
1489 /* Loop around if more data is required */
1490 if ((sz_copied
< buf_len
) && /* didn't get all requested data */
1491 (!skb_queue_empty(&sk
->sk_receive_queue
) ||
1492 (sz_copied
< target
)) && /* and more is ready or required */
1493 (!err
)) /* and haven't reached a FIN */
1498 return sz_copied
? sz_copied
: res
;
1502 * tipc_write_space - wake up thread if port congestion is released
1505 static void tipc_write_space(struct sock
*sk
)
1507 struct socket_wq
*wq
;
1510 wq
= rcu_dereference(sk
->sk_wq
);
1511 if (skwq_has_sleeper(wq
))
1512 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
1513 POLLWRNORM
| POLLWRBAND
);
1518 * tipc_data_ready - wake up threads to indicate messages have been received
1520 * @len: the length of messages
1522 static void tipc_data_ready(struct sock
*sk
)
1524 struct socket_wq
*wq
;
1527 wq
= rcu_dereference(sk
->sk_wq
);
1528 if (skwq_has_sleeper(wq
))
1529 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
1530 POLLRDNORM
| POLLRDBAND
);
1534 static void tipc_sock_destruct(struct sock
*sk
)
1536 __skb_queue_purge(&sk
->sk_receive_queue
);
1540 * filter_connect - Handle all incoming messages for a connection-based socket
1542 * @skb: pointer to message buffer. Set to NULL if buffer is consumed
1544 * Returns true if everything ok, false otherwise
1546 static bool filter_connect(struct tipc_sock
*tsk
, struct sk_buff
*skb
)
1548 struct sock
*sk
= &tsk
->sk
;
1549 struct net
*net
= sock_net(sk
);
1550 struct tipc_msg
*hdr
= buf_msg(skb
);
1552 if (unlikely(msg_mcast(hdr
)))
1555 switch (sk
->sk_state
) {
1556 case TIPC_CONNECTING
:
1557 /* Accept only ACK or NACK message */
1558 if (unlikely(!msg_connected(hdr
)))
1561 if (unlikely(msg_errcode(hdr
))) {
1562 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1563 sk
->sk_err
= ECONNREFUSED
;
1567 if (unlikely(!msg_isdata(hdr
))) {
1568 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1569 sk
->sk_err
= EINVAL
;
1573 tipc_sk_finish_conn(tsk
, msg_origport(hdr
), msg_orignode(hdr
));
1574 msg_set_importance(&tsk
->phdr
, msg_importance(hdr
));
1576 /* If 'ACK+' message, add to socket receive queue */
1577 if (msg_data_sz(hdr
))
1580 /* If empty 'ACK-' message, wake up sleeping connect() */
1581 if (waitqueue_active(sk_sleep(sk
)))
1582 wake_up_interruptible(sk_sleep(sk
));
1584 /* 'ACK-' message is neither accepted nor rejected: */
1585 msg_set_dest_droppable(hdr
, 1);
1589 case TIPC_DISCONNECTING
:
1592 /* Accept only SYN message */
1593 if (!msg_connected(hdr
) && !(msg_errcode(hdr
)))
1596 case TIPC_ESTABLISHED
:
1597 /* Accept only connection-based messages sent by peer */
1598 if (unlikely(!tsk_peer_msg(tsk
, hdr
)))
1601 if (unlikely(msg_errcode(hdr
))) {
1602 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
1603 /* Let timer expire on it's own */
1604 tipc_node_remove_conn(net
, tsk_peer_node(tsk
),
1606 sk
->sk_state_change(sk
);
1610 pr_err("Unknown sk_state %u\n", sk
->sk_state
);
1617 * rcvbuf_limit - get proper overload limit of socket receive queue
1621 * For connection oriented messages, irrespective of importance,
1622 * default queue limit is 2 MB.
1624 * For connectionless messages, queue limits are based on message
1625 * importance as follows:
1627 * TIPC_LOW_IMPORTANCE (2 MB)
1628 * TIPC_MEDIUM_IMPORTANCE (4 MB)
1629 * TIPC_HIGH_IMPORTANCE (8 MB)
1630 * TIPC_CRITICAL_IMPORTANCE (16 MB)
1632 * Returns overload limit according to corresponding message importance
1634 static unsigned int rcvbuf_limit(struct sock
*sk
, struct sk_buff
*skb
)
1636 struct tipc_sock
*tsk
= tipc_sk(sk
);
1637 struct tipc_msg
*hdr
= buf_msg(skb
);
1639 if (unlikely(!msg_connected(hdr
)))
1640 return sk
->sk_rcvbuf
<< msg_importance(hdr
);
1642 if (likely(tsk
->peer_caps
& TIPC_BLOCK_FLOWCTL
))
1643 return sk
->sk_rcvbuf
;
1645 return FLOWCTL_MSG_LIM
;
1649 * filter_rcv - validate incoming message
1651 * @skb: pointer to message.
1653 * Enqueues message on receive queue if acceptable; optionally handles
1654 * disconnect indication for a connected socket.
1656 * Called with socket lock already taken
1658 * Returns true if message was added to socket receive queue, otherwise false
1660 static bool filter_rcv(struct sock
*sk
, struct sk_buff
*skb
,
1661 struct sk_buff_head
*xmitq
)
1663 struct tipc_sock
*tsk
= tipc_sk(sk
);
1664 struct tipc_msg
*hdr
= buf_msg(skb
);
1665 unsigned int limit
= rcvbuf_limit(sk
, skb
);
1667 int usr
= msg_user(hdr
);
1670 if (unlikely(msg_user(hdr
) == CONN_MANAGER
)) {
1671 tipc_sk_proto_rcv(tsk
, skb
, xmitq
);
1675 if (unlikely(usr
== SOCK_WAKEUP
)) {
1676 onode
= msg_orignode(hdr
);
1678 u32_del(&tsk
->cong_links
, onode
);
1679 tsk
->cong_link_cnt
--;
1680 sk
->sk_write_space(sk
);
1684 /* Drop if illegal message type */
1685 if (unlikely(msg_type(hdr
) > TIPC_DIRECT_MSG
)) {
1690 /* Reject if wrong message type for current socket state */
1691 if (tipc_sk_type_connectionless(sk
)) {
1692 if (msg_connected(hdr
)) {
1693 err
= TIPC_ERR_NO_PORT
;
1696 } else if (unlikely(!filter_connect(tsk
, skb
))) {
1697 err
= TIPC_ERR_NO_PORT
;
1701 /* Reject message if there isn't room to queue it */
1702 if (unlikely(sk_rmem_alloc_get(sk
) + skb
->truesize
>= limit
)) {
1703 err
= TIPC_ERR_OVERLOAD
;
1707 /* Enqueue message */
1708 TIPC_SKB_CB(skb
)->bytes_read
= 0;
1709 __skb_queue_tail(&sk
->sk_receive_queue
, skb
);
1710 skb_set_owner_r(skb
, sk
);
1712 sk
->sk_data_ready(sk
);
1716 if (tipc_msg_reverse(tsk_own_node(tsk
), &skb
, err
))
1717 __skb_queue_tail(xmitq
, skb
);
1722 * tipc_backlog_rcv - handle incoming message from backlog queue
1726 * Caller must hold socket lock
1730 static int tipc_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1732 unsigned int truesize
= skb
->truesize
;
1733 struct sk_buff_head xmitq
;
1734 u32 dnode
, selector
;
1736 __skb_queue_head_init(&xmitq
);
1738 if (likely(filter_rcv(sk
, skb
, &xmitq
))) {
1739 atomic_add(truesize
, &tipc_sk(sk
)->dupl_rcvcnt
);
1743 if (skb_queue_empty(&xmitq
))
1746 /* Send response/rejected message */
1747 skb
= __skb_dequeue(&xmitq
);
1748 dnode
= msg_destnode(buf_msg(skb
));
1749 selector
= msg_origport(buf_msg(skb
));
1750 tipc_node_xmit_skb(sock_net(sk
), skb
, dnode
, selector
);
1755 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
1756 * inputq and try adding them to socket or backlog queue
1757 * @inputq: list of incoming buffers with potentially different destinations
1758 * @sk: socket where the buffers should be enqueued
1759 * @dport: port number for the socket
1761 * Caller must hold socket lock
1763 static void tipc_sk_enqueue(struct sk_buff_head
*inputq
, struct sock
*sk
,
1764 u32 dport
, struct sk_buff_head
*xmitq
)
1766 unsigned long time_limit
= jiffies
+ 2;
1767 struct sk_buff
*skb
;
1772 while (skb_queue_len(inputq
)) {
1773 if (unlikely(time_after_eq(jiffies
, time_limit
)))
1776 skb
= tipc_skb_dequeue(inputq
, dport
);
1780 /* Add message directly to receive queue if possible */
1781 if (!sock_owned_by_user(sk
)) {
1782 filter_rcv(sk
, skb
, xmitq
);
1786 /* Try backlog, compensating for double-counted bytes */
1787 dcnt
= &tipc_sk(sk
)->dupl_rcvcnt
;
1788 if (!sk
->sk_backlog
.len
)
1789 atomic_set(dcnt
, 0);
1790 lim
= rcvbuf_limit(sk
, skb
) + atomic_read(dcnt
);
1791 if (likely(!sk_add_backlog(sk
, skb
, lim
)))
1794 /* Overload => reject message back to sender */
1795 onode
= tipc_own_addr(sock_net(sk
));
1796 if (tipc_msg_reverse(onode
, &skb
, TIPC_ERR_OVERLOAD
))
1797 __skb_queue_tail(xmitq
, skb
);
1803 * tipc_sk_rcv - handle a chain of incoming buffers
1804 * @inputq: buffer list containing the buffers
1805 * Consumes all buffers in list until inputq is empty
1806 * Note: may be called in multiple threads referring to the same queue
1808 void tipc_sk_rcv(struct net
*net
, struct sk_buff_head
*inputq
)
1810 struct sk_buff_head xmitq
;
1811 u32 dnode
, dport
= 0;
1813 struct tipc_sock
*tsk
;
1815 struct sk_buff
*skb
;
1817 __skb_queue_head_init(&xmitq
);
1818 while (skb_queue_len(inputq
)) {
1819 dport
= tipc_skb_peek_port(inputq
, dport
);
1820 tsk
= tipc_sk_lookup(net
, dport
);
1824 if (likely(spin_trylock_bh(&sk
->sk_lock
.slock
))) {
1825 tipc_sk_enqueue(inputq
, sk
, dport
, &xmitq
);
1826 spin_unlock_bh(&sk
->sk_lock
.slock
);
1828 /* Send pending response/rejected messages, if any */
1829 while ((skb
= __skb_dequeue(&xmitq
))) {
1830 dnode
= msg_destnode(buf_msg(skb
));
1831 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
1837 /* No destination socket => dequeue skb if still there */
1838 skb
= tipc_skb_dequeue(inputq
, dport
);
1842 /* Try secondary lookup if unresolved named message */
1843 err
= TIPC_ERR_NO_PORT
;
1844 if (tipc_msg_lookup_dest(net
, skb
, &err
))
1847 /* Prepare for message rejection */
1848 if (!tipc_msg_reverse(tipc_own_addr(net
), &skb
, err
))
1851 dnode
= msg_destnode(buf_msg(skb
));
1852 tipc_node_xmit_skb(net
, skb
, dnode
, dport
);
1856 static int tipc_wait_for_connect(struct socket
*sock
, long *timeo_p
)
1858 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
1859 struct sock
*sk
= sock
->sk
;
1863 int err
= sock_error(sk
);
1868 if (signal_pending(current
))
1869 return sock_intr_errno(*timeo_p
);
1871 add_wait_queue(sk_sleep(sk
), &wait
);
1872 done
= sk_wait_event(sk
, timeo_p
,
1873 sk
->sk_state
!= TIPC_CONNECTING
, &wait
);
1874 remove_wait_queue(sk_sleep(sk
), &wait
);
1880 * tipc_connect - establish a connection to another TIPC port
1881 * @sock: socket structure
1882 * @dest: socket address for destination port
1883 * @destlen: size of socket address data structure
1884 * @flags: file-related flags associated with socket
1886 * Returns 0 on success, errno otherwise
1888 static int tipc_connect(struct socket
*sock
, struct sockaddr
*dest
,
1889 int destlen
, int flags
)
1891 struct sock
*sk
= sock
->sk
;
1892 struct tipc_sock
*tsk
= tipc_sk(sk
);
1893 struct sockaddr_tipc
*dst
= (struct sockaddr_tipc
*)dest
;
1894 struct msghdr m
= {NULL
,};
1895 long timeout
= (flags
& O_NONBLOCK
) ? 0 : tsk
->conn_timeout
;
1901 /* DGRAM/RDM connect(), just save the destaddr */
1902 if (tipc_sk_type_connectionless(sk
)) {
1903 if (dst
->family
== AF_UNSPEC
) {
1904 memset(&tsk
->peer
, 0, sizeof(struct sockaddr_tipc
));
1905 } else if (destlen
!= sizeof(struct sockaddr_tipc
)) {
1908 memcpy(&tsk
->peer
, dest
, destlen
);
1914 * Reject connection attempt using multicast address
1916 * Note: send_msg() validates the rest of the address fields,
1917 * so there's no need to do it here
1919 if (dst
->addrtype
== TIPC_ADDR_MCAST
) {
1924 previous
= sk
->sk_state
;
1926 switch (sk
->sk_state
) {
1928 /* Send a 'SYN-' to destination */
1930 m
.msg_namelen
= destlen
;
1932 /* If connect is in non-blocking case, set MSG_DONTWAIT to
1933 * indicate send_msg() is never blocked.
1936 m
.msg_flags
= MSG_DONTWAIT
;
1938 res
= __tipc_sendmsg(sock
, &m
, 0);
1939 if ((res
< 0) && (res
!= -EWOULDBLOCK
))
1942 /* Just entered TIPC_CONNECTING state; the only
1943 * difference is that return value in non-blocking
1944 * case is EINPROGRESS, rather than EALREADY.
1948 case TIPC_CONNECTING
:
1950 if (previous
== TIPC_CONNECTING
)
1954 timeout
= msecs_to_jiffies(timeout
);
1955 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
1956 res
= tipc_wait_for_connect(sock
, &timeout
);
1958 case TIPC_ESTABLISHED
:
1971 * tipc_listen - allow socket to listen for incoming connections
1972 * @sock: socket structure
1975 * Returns 0 on success, errno otherwise
1977 static int tipc_listen(struct socket
*sock
, int len
)
1979 struct sock
*sk
= sock
->sk
;
1983 res
= tipc_set_sk_state(sk
, TIPC_LISTEN
);
1989 static int tipc_wait_for_accept(struct socket
*sock
, long timeo
)
1991 struct sock
*sk
= sock
->sk
;
1995 /* True wake-one mechanism for incoming connections: only
1996 * one process gets woken up, not the 'whole herd'.
1997 * Since we do not 'race & poll' for established sockets
1998 * anymore, the common case will execute the loop only once.
2001 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
2002 TASK_INTERRUPTIBLE
);
2003 if (timeo
&& skb_queue_empty(&sk
->sk_receive_queue
)) {
2005 timeo
= schedule_timeout(timeo
);
2009 if (!skb_queue_empty(&sk
->sk_receive_queue
))
2014 err
= sock_intr_errno(timeo
);
2015 if (signal_pending(current
))
2018 finish_wait(sk_sleep(sk
), &wait
);
2023 * tipc_accept - wait for connection request
2024 * @sock: listening socket
2025 * @newsock: new socket that is to be connected
2026 * @flags: file-related flags associated with socket
2028 * Returns 0 on success, errno otherwise
2030 static int tipc_accept(struct socket
*sock
, struct socket
*new_sock
, int flags
)
2032 struct sock
*new_sk
, *sk
= sock
->sk
;
2033 struct sk_buff
*buf
;
2034 struct tipc_sock
*new_tsock
;
2035 struct tipc_msg
*msg
;
2041 if (sk
->sk_state
!= TIPC_LISTEN
) {
2045 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
2046 res
= tipc_wait_for_accept(sock
, timeo
);
2050 buf
= skb_peek(&sk
->sk_receive_queue
);
2052 res
= tipc_sk_create(sock_net(sock
->sk
), new_sock
, 0, 0);
2055 security_sk_clone(sock
->sk
, new_sock
->sk
);
2057 new_sk
= new_sock
->sk
;
2058 new_tsock
= tipc_sk(new_sk
);
2061 /* we lock on new_sk; but lockdep sees the lock on sk */
2062 lock_sock_nested(new_sk
, SINGLE_DEPTH_NESTING
);
2065 * Reject any stray messages received by new socket
2066 * before the socket lock was taken (very, very unlikely)
2068 tsk_rej_rx_queue(new_sk
);
2070 /* Connect new socket to it's peer */
2071 tipc_sk_finish_conn(new_tsock
, msg_origport(msg
), msg_orignode(msg
));
2073 tsk_set_importance(new_tsock
, msg_importance(msg
));
2074 if (msg_named(msg
)) {
2075 new_tsock
->conn_type
= msg_nametype(msg
);
2076 new_tsock
->conn_instance
= msg_nameinst(msg
);
2080 * Respond to 'SYN-' by discarding it & returning 'ACK'-.
2081 * Respond to 'SYN+' by queuing it on new socket.
2083 if (!msg_data_sz(msg
)) {
2084 struct msghdr m
= {NULL
,};
2086 tsk_advance_rx_queue(sk
);
2087 __tipc_sendstream(new_sock
, &m
, 0);
2089 __skb_dequeue(&sk
->sk_receive_queue
);
2090 __skb_queue_head(&new_sk
->sk_receive_queue
, buf
);
2091 skb_set_owner_r(buf
, new_sk
);
2093 release_sock(new_sk
);
2100 * tipc_shutdown - shutdown socket connection
2101 * @sock: socket structure
2102 * @how: direction to close (must be SHUT_RDWR)
2104 * Terminates connection (if necessary), then purges socket's receive queue.
2106 * Returns 0 on success, errno otherwise
2108 static int tipc_shutdown(struct socket
*sock
, int how
)
2110 struct sock
*sk
= sock
->sk
;
2113 if (how
!= SHUT_RDWR
)
2118 __tipc_shutdown(sock
, TIPC_CONN_SHUTDOWN
);
2119 sk
->sk_shutdown
= SEND_SHUTDOWN
;
2121 if (sk
->sk_state
== TIPC_DISCONNECTING
) {
2122 /* Discard any unreceived messages */
2123 __skb_queue_purge(&sk
->sk_receive_queue
);
2125 /* Wake up anyone sleeping in poll */
2126 sk
->sk_state_change(sk
);
2136 static void tipc_sk_timeout(unsigned long data
)
2138 struct tipc_sock
*tsk
= (struct tipc_sock
*)data
;
2139 struct sock
*sk
= &tsk
->sk
;
2140 struct sk_buff
*skb
= NULL
;
2141 u32 peer_port
, peer_node
;
2142 u32 own_node
= tsk_own_node(tsk
);
2145 if (!tipc_sk_connected(sk
)) {
2149 peer_port
= tsk_peer_port(tsk
);
2150 peer_node
= tsk_peer_node(tsk
);
2152 if (tsk
->probe_unacked
) {
2153 if (!sock_owned_by_user(sk
)) {
2154 tipc_set_sk_state(sk
, TIPC_DISCONNECTING
);
2155 tipc_node_remove_conn(sock_net(sk
), tsk_peer_node(tsk
),
2156 tsk_peer_port(tsk
));
2157 sk
->sk_state_change(sk
);
2159 /* Try again later */
2160 sk_reset_timer(sk
, &sk
->sk_timer
, (HZ
/ 20));
2167 skb
= tipc_msg_create(CONN_MANAGER
, CONN_PROBE
,
2168 INT_H_SIZE
, 0, peer_node
, own_node
,
2169 peer_port
, tsk
->portid
, TIPC_OK
);
2170 tsk
->probe_unacked
= true;
2171 sk_reset_timer(sk
, &sk
->sk_timer
, jiffies
+ CONN_PROBING_INTERVAL
);
2174 tipc_node_xmit_skb(sock_net(sk
), skb
, peer_node
, tsk
->portid
);
2179 static int tipc_sk_publish(struct tipc_sock
*tsk
, uint scope
,
2180 struct tipc_name_seq
const *seq
)
2182 struct sock
*sk
= &tsk
->sk
;
2183 struct net
*net
= sock_net(sk
);
2184 struct publication
*publ
;
2187 if (tipc_sk_connected(sk
))
2189 key
= tsk
->portid
+ tsk
->pub_count
+ 1;
2190 if (key
== tsk
->portid
)
2193 publ
= tipc_nametbl_publish(net
, seq
->type
, seq
->lower
, seq
->upper
,
2194 scope
, tsk
->portid
, key
);
2195 if (unlikely(!publ
))
2198 list_add(&publ
->pport_list
, &tsk
->publications
);
2204 static int tipc_sk_withdraw(struct tipc_sock
*tsk
, uint scope
,
2205 struct tipc_name_seq
const *seq
)
2207 struct net
*net
= sock_net(&tsk
->sk
);
2208 struct publication
*publ
;
2209 struct publication
*safe
;
2212 list_for_each_entry_safe(publ
, safe
, &tsk
->publications
, pport_list
) {
2214 if (publ
->scope
!= scope
)
2216 if (publ
->type
!= seq
->type
)
2218 if (publ
->lower
!= seq
->lower
)
2220 if (publ
->upper
!= seq
->upper
)
2222 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2223 publ
->ref
, publ
->key
);
2227 tipc_nametbl_withdraw(net
, publ
->type
, publ
->lower
,
2228 publ
->ref
, publ
->key
);
2231 if (list_empty(&tsk
->publications
))
2236 /* tipc_sk_reinit: set non-zero address in all existing sockets
2237 * when we go from standalone to network mode.
2239 void tipc_sk_reinit(struct net
*net
)
2241 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2242 struct rhashtable_iter iter
;
2243 struct tipc_sock
*tsk
;
2244 struct tipc_msg
*msg
;
2246 rhashtable_walk_enter(&tn
->sk_rht
, &iter
);
2249 tsk
= ERR_PTR(rhashtable_walk_start(&iter
));
2253 while ((tsk
= rhashtable_walk_next(&iter
)) && !IS_ERR(tsk
)) {
2254 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2256 msg_set_prevnode(msg
, tn
->own_addr
);
2257 msg_set_orignode(msg
, tn
->own_addr
);
2258 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2261 rhashtable_walk_stop(&iter
);
2262 } while (tsk
== ERR_PTR(-EAGAIN
));
2265 static struct tipc_sock
*tipc_sk_lookup(struct net
*net
, u32 portid
)
2267 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2268 struct tipc_sock
*tsk
;
2271 tsk
= rhashtable_lookup_fast(&tn
->sk_rht
, &portid
, tsk_rht_params
);
2273 sock_hold(&tsk
->sk
);
2279 static int tipc_sk_insert(struct tipc_sock
*tsk
)
2281 struct sock
*sk
= &tsk
->sk
;
2282 struct net
*net
= sock_net(sk
);
2283 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2284 u32 remaining
= (TIPC_MAX_PORT
- TIPC_MIN_PORT
) + 1;
2285 u32 portid
= prandom_u32() % remaining
+ TIPC_MIN_PORT
;
2287 while (remaining
--) {
2289 if ((portid
< TIPC_MIN_PORT
) || (portid
> TIPC_MAX_PORT
))
2290 portid
= TIPC_MIN_PORT
;
2291 tsk
->portid
= portid
;
2292 sock_hold(&tsk
->sk
);
2293 if (!rhashtable_lookup_insert_fast(&tn
->sk_rht
, &tsk
->node
,
2302 static void tipc_sk_remove(struct tipc_sock
*tsk
)
2304 struct sock
*sk
= &tsk
->sk
;
2305 struct tipc_net
*tn
= net_generic(sock_net(sk
), tipc_net_id
);
2307 if (!rhashtable_remove_fast(&tn
->sk_rht
, &tsk
->node
, tsk_rht_params
)) {
2308 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
2313 static const struct rhashtable_params tsk_rht_params
= {
2315 .head_offset
= offsetof(struct tipc_sock
, node
),
2316 .key_offset
= offsetof(struct tipc_sock
, portid
),
2317 .key_len
= sizeof(u32
), /* portid */
2318 .max_size
= 1048576,
2320 .automatic_shrinking
= true,
2323 int tipc_sk_rht_init(struct net
*net
)
2325 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2327 return rhashtable_init(&tn
->sk_rht
, &tsk_rht_params
);
2330 void tipc_sk_rht_destroy(struct net
*net
)
2332 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2334 /* Wait for socket readers to complete */
2337 rhashtable_destroy(&tn
->sk_rht
);
2341 * tipc_setsockopt - set socket option
2342 * @sock: socket structure
2343 * @lvl: option level
2344 * @opt: option identifier
2345 * @ov: pointer to new option value
2346 * @ol: length of option value
2348 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
2349 * (to ease compatibility).
2351 * Returns 0 on success, errno otherwise
2353 static int tipc_setsockopt(struct socket
*sock
, int lvl
, int opt
,
2354 char __user
*ov
, unsigned int ol
)
2356 struct sock
*sk
= sock
->sk
;
2357 struct tipc_sock
*tsk
= tipc_sk(sk
);
2361 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2363 if (lvl
!= SOL_TIPC
)
2364 return -ENOPROTOOPT
;
2367 case TIPC_IMPORTANCE
:
2368 case TIPC_SRC_DROPPABLE
:
2369 case TIPC_DEST_DROPPABLE
:
2370 case TIPC_CONN_TIMEOUT
:
2371 if (ol
< sizeof(value
))
2373 res
= get_user(value
, (u32 __user
*)ov
);
2385 case TIPC_IMPORTANCE
:
2386 res
= tsk_set_importance(tsk
, value
);
2388 case TIPC_SRC_DROPPABLE
:
2389 if (sock
->type
!= SOCK_STREAM
)
2390 tsk_set_unreliable(tsk
, value
);
2394 case TIPC_DEST_DROPPABLE
:
2395 tsk_set_unreturnable(tsk
, value
);
2397 case TIPC_CONN_TIMEOUT
:
2398 tipc_sk(sk
)->conn_timeout
= value
;
2400 case TIPC_MCAST_BROADCAST
:
2401 tsk
->mc_method
.rcast
= false;
2402 tsk
->mc_method
.mandatory
= true;
2404 case TIPC_MCAST_REPLICAST
:
2405 tsk
->mc_method
.rcast
= true;
2406 tsk
->mc_method
.mandatory
= true;
2418 * tipc_getsockopt - get socket option
2419 * @sock: socket structure
2420 * @lvl: option level
2421 * @opt: option identifier
2422 * @ov: receptacle for option value
2423 * @ol: receptacle for length of option value
2425 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
2426 * (to ease compatibility).
2428 * Returns 0 on success, errno otherwise
2430 static int tipc_getsockopt(struct socket
*sock
, int lvl
, int opt
,
2431 char __user
*ov
, int __user
*ol
)
2433 struct sock
*sk
= sock
->sk
;
2434 struct tipc_sock
*tsk
= tipc_sk(sk
);
2439 if ((lvl
== IPPROTO_TCP
) && (sock
->type
== SOCK_STREAM
))
2440 return put_user(0, ol
);
2441 if (lvl
!= SOL_TIPC
)
2442 return -ENOPROTOOPT
;
2443 res
= get_user(len
, ol
);
2450 case TIPC_IMPORTANCE
:
2451 value
= tsk_importance(tsk
);
2453 case TIPC_SRC_DROPPABLE
:
2454 value
= tsk_unreliable(tsk
);
2456 case TIPC_DEST_DROPPABLE
:
2457 value
= tsk_unreturnable(tsk
);
2459 case TIPC_CONN_TIMEOUT
:
2460 value
= tsk
->conn_timeout
;
2461 /* no need to set "res", since already 0 at this point */
2463 case TIPC_NODE_RECVQ_DEPTH
:
2464 value
= 0; /* was tipc_queue_size, now obsolete */
2466 case TIPC_SOCK_RECVQ_DEPTH
:
2467 value
= skb_queue_len(&sk
->sk_receive_queue
);
2476 return res
; /* "get" failed */
2478 if (len
< sizeof(value
))
2481 if (copy_to_user(ov
, &value
, sizeof(value
)))
2484 return put_user(sizeof(value
), ol
);
2487 static int tipc_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2489 struct sock
*sk
= sock
->sk
;
2490 struct tipc_sioc_ln_req lnr
;
2491 void __user
*argp
= (void __user
*)arg
;
2494 case SIOCGETLINKNAME
:
2495 if (copy_from_user(&lnr
, argp
, sizeof(lnr
)))
2497 if (!tipc_node_get_linkname(sock_net(sk
),
2498 lnr
.bearer_id
& 0xffff, lnr
.peer
,
2499 lnr
.linkname
, TIPC_MAX_LINK_NAME
)) {
2500 if (copy_to_user(argp
, &lnr
, sizeof(lnr
)))
2504 return -EADDRNOTAVAIL
;
2506 return -ENOIOCTLCMD
;
2510 /* Protocol switches for the various types of TIPC sockets */
2512 static const struct proto_ops msg_ops
= {
2513 .owner
= THIS_MODULE
,
2515 .release
= tipc_release
,
2517 .connect
= tipc_connect
,
2518 .socketpair
= sock_no_socketpair
,
2519 .accept
= sock_no_accept
,
2520 .getname
= tipc_getname
,
2522 .ioctl
= tipc_ioctl
,
2523 .listen
= sock_no_listen
,
2524 .shutdown
= tipc_shutdown
,
2525 .setsockopt
= tipc_setsockopt
,
2526 .getsockopt
= tipc_getsockopt
,
2527 .sendmsg
= tipc_sendmsg
,
2528 .recvmsg
= tipc_recvmsg
,
2529 .mmap
= sock_no_mmap
,
2530 .sendpage
= sock_no_sendpage
2533 static const struct proto_ops packet_ops
= {
2534 .owner
= THIS_MODULE
,
2536 .release
= tipc_release
,
2538 .connect
= tipc_connect
,
2539 .socketpair
= sock_no_socketpair
,
2540 .accept
= tipc_accept
,
2541 .getname
= tipc_getname
,
2543 .ioctl
= tipc_ioctl
,
2544 .listen
= tipc_listen
,
2545 .shutdown
= tipc_shutdown
,
2546 .setsockopt
= tipc_setsockopt
,
2547 .getsockopt
= tipc_getsockopt
,
2548 .sendmsg
= tipc_send_packet
,
2549 .recvmsg
= tipc_recvmsg
,
2550 .mmap
= sock_no_mmap
,
2551 .sendpage
= sock_no_sendpage
2554 static const struct proto_ops stream_ops
= {
2555 .owner
= THIS_MODULE
,
2557 .release
= tipc_release
,
2559 .connect
= tipc_connect
,
2560 .socketpair
= sock_no_socketpair
,
2561 .accept
= tipc_accept
,
2562 .getname
= tipc_getname
,
2564 .ioctl
= tipc_ioctl
,
2565 .listen
= tipc_listen
,
2566 .shutdown
= tipc_shutdown
,
2567 .setsockopt
= tipc_setsockopt
,
2568 .getsockopt
= tipc_getsockopt
,
2569 .sendmsg
= tipc_sendstream
,
2570 .recvmsg
= tipc_recv_stream
,
2571 .mmap
= sock_no_mmap
,
2572 .sendpage
= sock_no_sendpage
2575 static const struct net_proto_family tipc_family_ops
= {
2576 .owner
= THIS_MODULE
,
2578 .create
= tipc_sk_create
2581 static struct proto tipc_proto
= {
2583 .owner
= THIS_MODULE
,
2584 .obj_size
= sizeof(struct tipc_sock
),
2585 .sysctl_rmem
= sysctl_tipc_rmem
2589 * tipc_socket_init - initialize TIPC socket interface
2591 * Returns 0 on success, errno otherwise
2593 int tipc_socket_init(void)
2597 res
= proto_register(&tipc_proto
, 1);
2599 pr_err("Failed to register TIPC protocol type\n");
2603 res
= sock_register(&tipc_family_ops
);
2605 pr_err("Failed to register TIPC socket type\n");
2606 proto_unregister(&tipc_proto
);
2614 * tipc_socket_stop - stop TIPC socket interface
2616 void tipc_socket_stop(void)
2618 sock_unregister(tipc_family_ops
.family
);
2619 proto_unregister(&tipc_proto
);
2622 /* Caller should hold socket lock for the passed tipc socket. */
2623 static int __tipc_nl_add_sk_con(struct sk_buff
*skb
, struct tipc_sock
*tsk
)
2627 struct nlattr
*nest
;
2629 peer_node
= tsk_peer_node(tsk
);
2630 peer_port
= tsk_peer_port(tsk
);
2632 nest
= nla_nest_start(skb
, TIPC_NLA_SOCK_CON
);
2634 if (nla_put_u32(skb
, TIPC_NLA_CON_NODE
, peer_node
))
2636 if (nla_put_u32(skb
, TIPC_NLA_CON_SOCK
, peer_port
))
2639 if (tsk
->conn_type
!= 0) {
2640 if (nla_put_flag(skb
, TIPC_NLA_CON_FLAG
))
2642 if (nla_put_u32(skb
, TIPC_NLA_CON_TYPE
, tsk
->conn_type
))
2644 if (nla_put_u32(skb
, TIPC_NLA_CON_INST
, tsk
->conn_instance
))
2647 nla_nest_end(skb
, nest
);
2652 nla_nest_cancel(skb
, nest
);
2657 /* Caller should hold socket lock for the passed tipc socket. */
2658 static int __tipc_nl_add_sk(struct sk_buff
*skb
, struct netlink_callback
*cb
,
2659 struct tipc_sock
*tsk
)
2663 struct nlattr
*attrs
;
2664 struct net
*net
= sock_net(skb
->sk
);
2665 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2666 struct sock
*sk
= &tsk
->sk
;
2668 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2669 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_SOCK_GET
);
2673 attrs
= nla_nest_start(skb
, TIPC_NLA_SOCK
);
2675 goto genlmsg_cancel
;
2676 if (nla_put_u32(skb
, TIPC_NLA_SOCK_REF
, tsk
->portid
))
2677 goto attr_msg_cancel
;
2678 if (nla_put_u32(skb
, TIPC_NLA_SOCK_ADDR
, tn
->own_addr
))
2679 goto attr_msg_cancel
;
2681 if (tipc_sk_connected(sk
)) {
2682 err
= __tipc_nl_add_sk_con(skb
, tsk
);
2684 goto attr_msg_cancel
;
2685 } else if (!list_empty(&tsk
->publications
)) {
2686 if (nla_put_flag(skb
, TIPC_NLA_SOCK_HAS_PUBL
))
2687 goto attr_msg_cancel
;
2689 nla_nest_end(skb
, attrs
);
2690 genlmsg_end(skb
, hdr
);
2695 nla_nest_cancel(skb
, attrs
);
2697 genlmsg_cancel(skb
, hdr
);
2702 int tipc_nl_sk_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2705 struct tipc_sock
*tsk
;
2706 const struct bucket_table
*tbl
;
2707 struct rhash_head
*pos
;
2708 struct net
*net
= sock_net(skb
->sk
);
2709 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
2710 u32 tbl_id
= cb
->args
[0];
2711 u32 prev_portid
= cb
->args
[1];
2714 tbl
= rht_dereference_rcu((&tn
->sk_rht
)->tbl
, &tn
->sk_rht
);
2715 for (; tbl_id
< tbl
->size
; tbl_id
++) {
2716 rht_for_each_entry_rcu(tsk
, pos
, tbl
, tbl_id
, node
) {
2717 spin_lock_bh(&tsk
->sk
.sk_lock
.slock
);
2718 if (prev_portid
&& prev_portid
!= tsk
->portid
) {
2719 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2723 err
= __tipc_nl_add_sk(skb
, cb
, tsk
);
2725 prev_portid
= tsk
->portid
;
2726 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2730 spin_unlock_bh(&tsk
->sk
.sk_lock
.slock
);
2735 cb
->args
[0] = tbl_id
;
2736 cb
->args
[1] = prev_portid
;
2741 /* Caller should hold socket lock for the passed tipc socket. */
2742 static int __tipc_nl_add_sk_publ(struct sk_buff
*skb
,
2743 struct netlink_callback
*cb
,
2744 struct publication
*publ
)
2747 struct nlattr
*attrs
;
2749 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2750 &tipc_genl_family
, NLM_F_MULTI
, TIPC_NL_PUBL_GET
);
2754 attrs
= nla_nest_start(skb
, TIPC_NLA_PUBL
);
2756 goto genlmsg_cancel
;
2758 if (nla_put_u32(skb
, TIPC_NLA_PUBL_KEY
, publ
->key
))
2759 goto attr_msg_cancel
;
2760 if (nla_put_u32(skb
, TIPC_NLA_PUBL_TYPE
, publ
->type
))
2761 goto attr_msg_cancel
;
2762 if (nla_put_u32(skb
, TIPC_NLA_PUBL_LOWER
, publ
->lower
))
2763 goto attr_msg_cancel
;
2764 if (nla_put_u32(skb
, TIPC_NLA_PUBL_UPPER
, publ
->upper
))
2765 goto attr_msg_cancel
;
2767 nla_nest_end(skb
, attrs
);
2768 genlmsg_end(skb
, hdr
);
2773 nla_nest_cancel(skb
, attrs
);
2775 genlmsg_cancel(skb
, hdr
);
2780 /* Caller should hold socket lock for the passed tipc socket. */
2781 static int __tipc_nl_list_sk_publ(struct sk_buff
*skb
,
2782 struct netlink_callback
*cb
,
2783 struct tipc_sock
*tsk
, u32
*last_publ
)
2786 struct publication
*p
;
2789 list_for_each_entry(p
, &tsk
->publications
, pport_list
) {
2790 if (p
->key
== *last_publ
)
2793 if (p
->key
!= *last_publ
) {
2794 /* We never set seq or call nl_dump_check_consistent()
2795 * this means that setting prev_seq here will cause the
2796 * consistence check to fail in the netlink callback
2797 * handler. Resulting in the last NLMSG_DONE message
2798 * having the NLM_F_DUMP_INTR flag set.
2805 p
= list_first_entry(&tsk
->publications
, struct publication
,
2809 list_for_each_entry_from(p
, &tsk
->publications
, pport_list
) {
2810 err
= __tipc_nl_add_sk_publ(skb
, cb
, p
);
2812 *last_publ
= p
->key
;
2821 int tipc_nl_publ_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2824 u32 tsk_portid
= cb
->args
[0];
2825 u32 last_publ
= cb
->args
[1];
2826 u32 done
= cb
->args
[2];
2827 struct net
*net
= sock_net(skb
->sk
);
2828 struct tipc_sock
*tsk
;
2831 struct nlattr
**attrs
;
2832 struct nlattr
*sock
[TIPC_NLA_SOCK_MAX
+ 1];
2834 err
= tipc_nlmsg_parse(cb
->nlh
, &attrs
);
2838 if (!attrs
[TIPC_NLA_SOCK
])
2841 err
= nla_parse_nested(sock
, TIPC_NLA_SOCK_MAX
,
2842 attrs
[TIPC_NLA_SOCK
],
2843 tipc_nl_sock_policy
);
2847 if (!sock
[TIPC_NLA_SOCK_REF
])
2850 tsk_portid
= nla_get_u32(sock
[TIPC_NLA_SOCK_REF
]);
2856 tsk
= tipc_sk_lookup(net
, tsk_portid
);
2860 lock_sock(&tsk
->sk
);
2861 err
= __tipc_nl_list_sk_publ(skb
, cb
, tsk
, &last_publ
);
2864 release_sock(&tsk
->sk
);
2867 cb
->args
[0] = tsk_portid
;
2868 cb
->args
[1] = last_publ
;