2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 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.
42 #include "name_distr.h"
46 #include <linux/pkt_sched.h>
49 u32 sent_info
; /* used in counting # sent packets */
50 u32 recv_info
; /* used in counting # recv'd packets */
67 u32 link_congs
; /* # port sends blocked by congestion */
70 u32 max_queue_sz
; /* send queue size high water mark */
71 u32 accu_queue_sz
; /* used for send queue size profiling */
72 u32 queue_sz_counts
; /* used for send queue size profiling */
73 u32 msg_length_counts
; /* used for message length profiling */
74 u32 msg_lengths_total
; /* used for message length profiling */
75 u32 msg_length_profile
[7]; /* used for msg. length profiling */
79 * struct tipc_link - TIPC link data structure
80 * @addr: network address of link's peer node
81 * @name: link name character string
82 * @media_addr: media address to use when sending messages over link
84 * @net: pointer to namespace struct
85 * @refcnt: reference counter for permanent references (owner node & timer)
86 * @peer_session: link session # being used by peer end of link
87 * @peer_bearer_id: bearer id used by link's peer endpoint
88 * @bearer_id: local bearer id used by link
89 * @tolerance: minimum link continuity loss needed to reset link [in ms]
90 * @keepalive_intv: link keepalive timer interval
91 * @abort_limit: # of unacknowledged continuity probes needed to reset link
92 * @state: current state of link FSM
93 * @peer_caps: bitmap describing capabilities of peer node
94 * @silent_intv_cnt: # of timer intervals without any reception from peer
95 * @proto_msg: template for control messages generated by link
96 * @pmsg: convenience pointer to "proto_msg" field
97 * @priority: current link priority
98 * @net_plane: current link network plane ('A' through 'H')
99 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
100 * @exp_msg_count: # of tunnelled messages expected during link changeover
101 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
102 * @mtu: current maximum packet size for this link
103 * @advertised_mtu: advertised own mtu when link is being established
104 * @transmitq: queue for sent, non-acked messages
105 * @backlogq: queue for messages waiting to be sent
106 * @snt_nxt: next sequence number to use for outbound messages
107 * @last_retransmitted: sequence number of most recently retransmitted message
108 * @stale_count: # of identical retransmit requests made by peer
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
125 char name
[TIPC_MAX_LINK_NAME
];
128 /* Management and link supervision data */
134 unsigned long keepalive_intv
;
140 char if_name
[TIPC_MAX_IF_NAME
];
147 struct sk_buff
*failover_reasm_skb
;
149 /* Max packet negotiation */
154 struct sk_buff_head transmq
;
155 struct sk_buff_head backlogq
;
168 struct sk_buff_head deferdq
;
169 struct sk_buff_head
*inputq
;
170 struct sk_buff_head
*namedq
;
172 /* Congestion handling */
173 struct sk_buff_head wakeupq
;
175 /* Fragmentation/reassembly */
176 struct sk_buff
*reasm_buf
;
181 struct tipc_link
*bc_rcvlink
;
182 struct tipc_link
*bc_sndlink
;
187 struct tipc_stats stats
;
191 * Error message prefixes
193 static const char *link_co_err
= "Link tunneling error, ";
194 static const char *link_rst_msg
= "Resetting link ";
196 /* Send states for broadcast NACKs
199 BC_NACK_SND_CONDITIONAL
,
200 BC_NACK_SND_UNCONDITIONAL
,
201 BC_NACK_SND_SUPPRESS
,
205 * Interval between NACKs when packets arrive out of order
207 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
209 /* Wildcard value for link session numbers. When it is known that
210 * peer endpoint is down, any session number must be accepted.
212 #define ANY_SESSION 0x10000
217 LINK_ESTABLISHED
= 0xe,
218 LINK_ESTABLISHING
= 0xe << 4,
219 LINK_RESET
= 0x1 << 8,
220 LINK_RESETTING
= 0x2 << 12,
221 LINK_PEER_RESET
= 0xd << 16,
222 LINK_FAILINGOVER
= 0xf << 20,
223 LINK_SYNCHING
= 0xc << 24
226 /* Link FSM state checking routines
228 static int link_is_up(struct tipc_link
*l
)
230 return l
->state
& (LINK_ESTABLISHED
| LINK_SYNCHING
);
233 static int tipc_link_proto_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
234 struct sk_buff_head
*xmitq
);
235 static void tipc_link_build_proto_msg(struct tipc_link
*l
, int mtyp
, bool probe
,
236 u16 rcvgap
, int tolerance
, int priority
,
237 struct sk_buff_head
*xmitq
);
238 static void link_print(struct tipc_link
*l
, const char *str
);
239 static void tipc_link_build_nack_msg(struct tipc_link
*l
,
240 struct sk_buff_head
*xmitq
);
241 static void tipc_link_build_bc_init_msg(struct tipc_link
*l
,
242 struct sk_buff_head
*xmitq
);
243 static bool tipc_link_release_pkts(struct tipc_link
*l
, u16 to
);
246 * Simple non-static link routines (i.e. referenced outside this file)
248 bool tipc_link_is_up(struct tipc_link
*l
)
250 return link_is_up(l
);
253 bool tipc_link_peer_is_down(struct tipc_link
*l
)
255 return l
->state
== LINK_PEER_RESET
;
258 bool tipc_link_is_reset(struct tipc_link
*l
)
260 return l
->state
& (LINK_RESET
| LINK_FAILINGOVER
| LINK_ESTABLISHING
);
263 bool tipc_link_is_establishing(struct tipc_link
*l
)
265 return l
->state
== LINK_ESTABLISHING
;
268 bool tipc_link_is_synching(struct tipc_link
*l
)
270 return l
->state
== LINK_SYNCHING
;
273 bool tipc_link_is_failingover(struct tipc_link
*l
)
275 return l
->state
== LINK_FAILINGOVER
;
278 bool tipc_link_is_blocked(struct tipc_link
*l
)
280 return l
->state
& (LINK_RESETTING
| LINK_PEER_RESET
| LINK_FAILINGOVER
);
283 static bool link_is_bc_sndlink(struct tipc_link
*l
)
285 return !l
->bc_sndlink
;
288 static bool link_is_bc_rcvlink(struct tipc_link
*l
)
290 return ((l
->bc_rcvlink
== l
) && !link_is_bc_sndlink(l
));
293 int tipc_link_is_active(struct tipc_link
*l
)
298 void tipc_link_set_active(struct tipc_link
*l
, bool active
)
303 u32
tipc_link_id(struct tipc_link
*l
)
305 return l
->peer_bearer_id
<< 16 | l
->bearer_id
;
308 int tipc_link_window(struct tipc_link
*l
)
313 int tipc_link_prio(struct tipc_link
*l
)
318 unsigned long tipc_link_tolerance(struct tipc_link
*l
)
323 struct sk_buff_head
*tipc_link_inputq(struct tipc_link
*l
)
328 char tipc_link_plane(struct tipc_link
*l
)
333 void tipc_link_add_bc_peer(struct tipc_link
*snd_l
,
334 struct tipc_link
*uc_l
,
335 struct sk_buff_head
*xmitq
)
337 struct tipc_link
*rcv_l
= uc_l
->bc_rcvlink
;
340 rcv_l
->acked
= snd_l
->snd_nxt
- 1;
341 snd_l
->state
= LINK_ESTABLISHED
;
342 tipc_link_build_bc_init_msg(uc_l
, xmitq
);
345 void tipc_link_remove_bc_peer(struct tipc_link
*snd_l
,
346 struct tipc_link
*rcv_l
,
347 struct sk_buff_head
*xmitq
)
349 u16 ack
= snd_l
->snd_nxt
- 1;
352 tipc_link_bc_ack_rcv(rcv_l
, ack
, xmitq
);
353 tipc_link_reset(rcv_l
);
354 rcv_l
->state
= LINK_RESET
;
355 if (!snd_l
->ackers
) {
356 tipc_link_reset(snd_l
);
357 snd_l
->state
= LINK_RESET
;
358 __skb_queue_purge(xmitq
);
362 int tipc_link_bc_peers(struct tipc_link
*l
)
367 void tipc_link_set_mtu(struct tipc_link
*l
, int mtu
)
372 int tipc_link_mtu(struct tipc_link
*l
)
377 u16
tipc_link_rcv_nxt(struct tipc_link
*l
)
382 u16
tipc_link_acked(struct tipc_link
*l
)
387 char *tipc_link_name(struct tipc_link
*l
)
393 * tipc_link_create - create a new link
394 * @n: pointer to associated node
395 * @if_name: associated interface name
396 * @bearer_id: id (index) of associated bearer
397 * @tolerance: link tolerance to be used by link
398 * @net_plane: network plane (A,B,c..) this link belongs to
399 * @mtu: mtu to be advertised by link
400 * @priority: priority to be used by link
401 * @window: send window to be used by link
402 * @session: session to be used by link
403 * @ownnode: identity of own node
404 * @peer: node id of peer node
405 * @peer_caps: bitmap describing peer node capabilities
406 * @bc_sndlink: the namespace global link used for broadcast sending
407 * @bc_rcvlink: the peer specific link used for broadcast reception
408 * @inputq: queue to put messages ready for delivery
409 * @namedq: queue to put binding table update messages ready for delivery
410 * @link: return value, pointer to put the created link
412 * Returns true if link was created, otherwise false
414 bool tipc_link_create(struct net
*net
, char *if_name
, int bearer_id
,
415 int tolerance
, char net_plane
, u32 mtu
, int priority
,
416 int window
, u32 session
, u32 ownnode
, u32 peer
,
418 struct tipc_link
*bc_sndlink
,
419 struct tipc_link
*bc_rcvlink
,
420 struct sk_buff_head
*inputq
,
421 struct sk_buff_head
*namedq
,
422 struct tipc_link
**link
)
426 l
= kzalloc(sizeof(*l
), GFP_ATOMIC
);
430 l
->session
= session
;
432 /* Note: peer i/f name is completed by reset/activate message */
433 sprintf(l
->name
, "%u.%u.%u:%s-%u.%u.%u:unknown",
434 tipc_zone(ownnode
), tipc_cluster(ownnode
), tipc_node(ownnode
),
435 if_name
, tipc_zone(peer
), tipc_cluster(peer
), tipc_node(peer
));
436 strcpy(l
->if_name
, if_name
);
438 l
->peer_caps
= peer_caps
;
440 l
->peer_session
= ANY_SESSION
;
441 l
->bearer_id
= bearer_id
;
442 l
->tolerance
= tolerance
;
443 l
->net_plane
= net_plane
;
444 l
->advertised_mtu
= mtu
;
446 l
->priority
= priority
;
447 tipc_link_set_queue_limits(l
, window
);
449 l
->bc_sndlink
= bc_sndlink
;
450 l
->bc_rcvlink
= bc_rcvlink
;
453 l
->state
= LINK_RESETTING
;
454 __skb_queue_head_init(&l
->transmq
);
455 __skb_queue_head_init(&l
->backlogq
);
456 __skb_queue_head_init(&l
->deferdq
);
457 skb_queue_head_init(&l
->wakeupq
);
458 skb_queue_head_init(l
->inputq
);
463 * tipc_link_bc_create - create new link to be used for broadcast
464 * @n: pointer to associated node
465 * @mtu: mtu to be used
466 * @window: send window to be used
467 * @inputq: queue to put messages ready for delivery
468 * @namedq: queue to put binding table update messages ready for delivery
469 * @link: return value, pointer to put the created link
471 * Returns true if link was created, otherwise false
473 bool tipc_link_bc_create(struct net
*net
, u32 ownnode
, u32 peer
,
474 int mtu
, int window
, u16 peer_caps
,
475 struct sk_buff_head
*inputq
,
476 struct sk_buff_head
*namedq
,
477 struct tipc_link
*bc_sndlink
,
478 struct tipc_link
**link
)
482 if (!tipc_link_create(net
, "", MAX_BEARERS
, 0, 'Z', mtu
, 0, window
,
483 0, ownnode
, peer
, peer_caps
, bc_sndlink
,
484 NULL
, inputq
, namedq
, link
))
488 strcpy(l
->name
, tipc_bclink_name
);
490 l
->state
= LINK_RESET
;
494 /* Broadcast send link is always up */
495 if (link_is_bc_sndlink(l
))
496 l
->state
= LINK_ESTABLISHED
;
502 * tipc_link_fsm_evt - link finite state machine
503 * @l: pointer to link
504 * @evt: state machine event to be processed
506 int tipc_link_fsm_evt(struct tipc_link
*l
, int evt
)
513 case LINK_PEER_RESET_EVT
:
514 l
->state
= LINK_PEER_RESET
;
517 l
->state
= LINK_RESET
;
519 case LINK_FAILURE_EVT
:
520 case LINK_FAILOVER_BEGIN_EVT
:
521 case LINK_ESTABLISH_EVT
:
522 case LINK_FAILOVER_END_EVT
:
523 case LINK_SYNCH_BEGIN_EVT
:
524 case LINK_SYNCH_END_EVT
:
531 case LINK_PEER_RESET_EVT
:
532 l
->state
= LINK_ESTABLISHING
;
534 case LINK_FAILOVER_BEGIN_EVT
:
535 l
->state
= LINK_FAILINGOVER
;
536 case LINK_FAILURE_EVT
:
538 case LINK_ESTABLISH_EVT
:
539 case LINK_FAILOVER_END_EVT
:
541 case LINK_SYNCH_BEGIN_EVT
:
542 case LINK_SYNCH_END_EVT
:
547 case LINK_PEER_RESET
:
550 l
->state
= LINK_ESTABLISHING
;
552 case LINK_PEER_RESET_EVT
:
553 case LINK_ESTABLISH_EVT
:
554 case LINK_FAILURE_EVT
:
556 case LINK_SYNCH_BEGIN_EVT
:
557 case LINK_SYNCH_END_EVT
:
558 case LINK_FAILOVER_BEGIN_EVT
:
559 case LINK_FAILOVER_END_EVT
:
564 case LINK_FAILINGOVER
:
566 case LINK_FAILOVER_END_EVT
:
567 l
->state
= LINK_RESET
;
569 case LINK_PEER_RESET_EVT
:
571 case LINK_ESTABLISH_EVT
:
572 case LINK_FAILURE_EVT
:
574 case LINK_FAILOVER_BEGIN_EVT
:
575 case LINK_SYNCH_BEGIN_EVT
:
576 case LINK_SYNCH_END_EVT
:
581 case LINK_ESTABLISHING
:
583 case LINK_ESTABLISH_EVT
:
584 l
->state
= LINK_ESTABLISHED
;
586 case LINK_FAILOVER_BEGIN_EVT
:
587 l
->state
= LINK_FAILINGOVER
;
590 l
->state
= LINK_RESET
;
592 case LINK_FAILURE_EVT
:
593 case LINK_PEER_RESET_EVT
:
594 case LINK_SYNCH_BEGIN_EVT
:
595 case LINK_FAILOVER_END_EVT
:
597 case LINK_SYNCH_END_EVT
:
602 case LINK_ESTABLISHED
:
604 case LINK_PEER_RESET_EVT
:
605 l
->state
= LINK_PEER_RESET
;
606 rc
|= TIPC_LINK_DOWN_EVT
;
608 case LINK_FAILURE_EVT
:
609 l
->state
= LINK_RESETTING
;
610 rc
|= TIPC_LINK_DOWN_EVT
;
613 l
->state
= LINK_RESET
;
615 case LINK_ESTABLISH_EVT
:
616 case LINK_SYNCH_END_EVT
:
618 case LINK_SYNCH_BEGIN_EVT
:
619 l
->state
= LINK_SYNCHING
;
621 case LINK_FAILOVER_BEGIN_EVT
:
622 case LINK_FAILOVER_END_EVT
:
629 case LINK_PEER_RESET_EVT
:
630 l
->state
= LINK_PEER_RESET
;
631 rc
|= TIPC_LINK_DOWN_EVT
;
633 case LINK_FAILURE_EVT
:
634 l
->state
= LINK_RESETTING
;
635 rc
|= TIPC_LINK_DOWN_EVT
;
638 l
->state
= LINK_RESET
;
640 case LINK_ESTABLISH_EVT
:
641 case LINK_SYNCH_BEGIN_EVT
:
643 case LINK_SYNCH_END_EVT
:
644 l
->state
= LINK_ESTABLISHED
;
646 case LINK_FAILOVER_BEGIN_EVT
:
647 case LINK_FAILOVER_END_EVT
:
653 pr_err("Unknown FSM state %x in %s\n", l
->state
, l
->name
);
657 pr_err("Illegal FSM event %x in state %x on link %s\n",
658 evt
, l
->state
, l
->name
);
662 /* link_profile_stats - update statistical profiling of traffic
664 static void link_profile_stats(struct tipc_link
*l
)
667 struct tipc_msg
*msg
;
670 /* Update counters used in statistical profiling of send traffic */
671 l
->stats
.accu_queue_sz
+= skb_queue_len(&l
->transmq
);
672 l
->stats
.queue_sz_counts
++;
674 skb
= skb_peek(&l
->transmq
);
678 length
= msg_size(msg
);
680 if (msg_user(msg
) == MSG_FRAGMENTER
) {
681 if (msg_type(msg
) != FIRST_FRAGMENT
)
683 length
= msg_size(msg_get_wrapped(msg
));
685 l
->stats
.msg_lengths_total
+= length
;
686 l
->stats
.msg_length_counts
++;
688 l
->stats
.msg_length_profile
[0]++;
689 else if (length
<= 256)
690 l
->stats
.msg_length_profile
[1]++;
691 else if (length
<= 1024)
692 l
->stats
.msg_length_profile
[2]++;
693 else if (length
<= 4096)
694 l
->stats
.msg_length_profile
[3]++;
695 else if (length
<= 16384)
696 l
->stats
.msg_length_profile
[4]++;
697 else if (length
<= 32768)
698 l
->stats
.msg_length_profile
[5]++;
700 l
->stats
.msg_length_profile
[6]++;
703 /* tipc_link_timeout - perform periodic task as instructed from node timeout
705 int tipc_link_timeout(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
712 u16 bc_snt
= l
->bc_sndlink
->snd_nxt
- 1;
713 u16 bc_acked
= l
->bc_rcvlink
->acked
;
715 link_profile_stats(l
);
718 case LINK_ESTABLISHED
:
720 if (l
->silent_intv_cnt
> l
->abort_limit
)
721 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
723 state
= bc_acked
!= bc_snt
;
724 probe
= l
->silent_intv_cnt
;
725 l
->silent_intv_cnt
++;
728 setup
= l
->rst_cnt
++ <= 4;
729 setup
|= !(l
->rst_cnt
% 16);
732 case LINK_ESTABLISHING
:
736 case LINK_PEER_RESET
:
738 case LINK_FAILINGOVER
:
744 if (state
|| probe
|| setup
)
745 tipc_link_build_proto_msg(l
, mtyp
, probe
, 0, 0, 0, xmitq
);
751 * link_schedule_user - schedule a message sender for wakeup after congestion
752 * @link: congested link
753 * @list: message that was attempted sent
754 * Create pseudo msg to send back to user when congestion abates
755 * Does not consume buffer list
757 static int link_schedule_user(struct tipc_link
*link
, struct sk_buff_head
*list
)
759 struct tipc_msg
*msg
= buf_msg(skb_peek(list
));
760 int imp
= msg_importance(msg
);
761 u32 oport
= msg_origport(msg
);
762 u32 addr
= tipc_own_addr(link
->net
);
765 /* This really cannot happen... */
766 if (unlikely(imp
> TIPC_CRITICAL_IMPORTANCE
)) {
767 pr_warn("%s<%s>, send queue full", link_rst_msg
, link
->name
);
770 /* Non-blocking sender: */
771 if (TIPC_SKB_CB(skb_peek(list
))->wakeup_pending
)
774 /* Create and schedule wakeup pseudo message */
775 skb
= tipc_msg_create(SOCK_WAKEUP
, 0, INT_H_SIZE
, 0,
776 addr
, addr
, oport
, 0, 0);
779 TIPC_SKB_CB(skb
)->chain_sz
= skb_queue_len(list
);
780 TIPC_SKB_CB(skb
)->chain_imp
= imp
;
781 skb_queue_tail(&link
->wakeupq
, skb
);
782 link
->stats
.link_congs
++;
787 * link_prepare_wakeup - prepare users for wakeup after congestion
788 * @link: congested link
789 * Move a number of waiting users, as permitted by available space in
790 * the send queue, from link wait queue to node wait queue for wakeup
792 void link_prepare_wakeup(struct tipc_link
*l
)
794 int pnd
[TIPC_SYSTEM_IMPORTANCE
+ 1] = {0,};
796 struct sk_buff
*skb
, *tmp
;
798 skb_queue_walk_safe(&l
->wakeupq
, skb
, tmp
) {
799 imp
= TIPC_SKB_CB(skb
)->chain_imp
;
800 lim
= l
->window
+ l
->backlog
[imp
].limit
;
801 pnd
[imp
] += TIPC_SKB_CB(skb
)->chain_sz
;
802 if ((pnd
[imp
] + l
->backlog
[imp
].len
) >= lim
)
804 skb_unlink(skb
, &l
->wakeupq
);
805 skb_queue_tail(l
->inputq
, skb
);
809 void tipc_link_reset(struct tipc_link
*l
)
811 l
->peer_session
= ANY_SESSION
;
813 l
->mtu
= l
->advertised_mtu
;
814 __skb_queue_purge(&l
->transmq
);
815 __skb_queue_purge(&l
->deferdq
);
816 skb_queue_splice_init(&l
->wakeupq
, l
->inputq
);
817 __skb_queue_purge(&l
->backlogq
);
818 l
->backlog
[TIPC_LOW_IMPORTANCE
].len
= 0;
819 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].len
= 0;
820 l
->backlog
[TIPC_HIGH_IMPORTANCE
].len
= 0;
821 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].len
= 0;
822 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].len
= 0;
823 kfree_skb(l
->reasm_buf
);
824 kfree_skb(l
->failover_reasm_skb
);
826 l
->failover_reasm_skb
= NULL
;
831 l
->silent_intv_cnt
= 0;
833 l
->stats
.recv_info
= 0;
835 l
->bc_peer_is_up
= false;
836 tipc_link_reset_stats(l
);
840 * tipc_link_xmit(): enqueue buffer list according to queue situation
842 * @list: chain of buffers containing message
843 * @xmitq: returned list of packets to be sent by caller
845 * Consumes the buffer chain, except when returning -ELINKCONG,
846 * since the caller then may want to make more send attempts.
847 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
848 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
850 int tipc_link_xmit(struct tipc_link
*l
, struct sk_buff_head
*list
,
851 struct sk_buff_head
*xmitq
)
853 struct tipc_msg
*hdr
= buf_msg(skb_peek(list
));
854 unsigned int maxwin
= l
->window
;
855 unsigned int i
, imp
= msg_importance(hdr
);
856 unsigned int mtu
= l
->mtu
;
857 u16 ack
= l
->rcv_nxt
- 1;
858 u16 seqno
= l
->snd_nxt
;
859 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
860 struct sk_buff_head
*transmq
= &l
->transmq
;
861 struct sk_buff_head
*backlogq
= &l
->backlogq
;
862 struct sk_buff
*skb
, *_skb
, *bskb
;
864 /* Match msg importance against this and all higher backlog limits: */
865 for (i
= imp
; i
<= TIPC_SYSTEM_IMPORTANCE
; i
++) {
866 if (unlikely(l
->backlog
[i
].len
>= l
->backlog
[i
].limit
))
867 return link_schedule_user(l
, list
);
869 if (unlikely(msg_size(hdr
) > mtu
)) {
870 skb_queue_purge(list
);
874 /* Prepare each packet for sending, and add to relevant queue: */
875 while (skb_queue_len(list
)) {
876 skb
= skb_peek(list
);
878 msg_set_seqno(hdr
, seqno
);
879 msg_set_ack(hdr
, ack
);
880 msg_set_bcast_ack(hdr
, bc_ack
);
882 if (likely(skb_queue_len(transmq
) < maxwin
)) {
883 _skb
= skb_clone(skb
, GFP_ATOMIC
);
885 skb_queue_purge(list
);
889 __skb_queue_tail(transmq
, skb
);
890 __skb_queue_tail(xmitq
, _skb
);
891 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
896 if (tipc_msg_bundle(skb_peek_tail(backlogq
), hdr
, mtu
)) {
897 kfree_skb(__skb_dequeue(list
));
898 l
->stats
.sent_bundled
++;
901 if (tipc_msg_make_bundle(&bskb
, hdr
, mtu
, l
->addr
)) {
902 kfree_skb(__skb_dequeue(list
));
903 __skb_queue_tail(backlogq
, bskb
);
904 l
->backlog
[msg_importance(buf_msg(bskb
))].len
++;
905 l
->stats
.sent_bundled
++;
906 l
->stats
.sent_bundles
++;
909 l
->backlog
[imp
].len
+= skb_queue_len(list
);
910 skb_queue_splice_tail_init(list
, backlogq
);
916 void tipc_link_advance_backlog(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
918 struct sk_buff
*skb
, *_skb
;
919 struct tipc_msg
*hdr
;
920 u16 seqno
= l
->snd_nxt
;
921 u16 ack
= l
->rcv_nxt
- 1;
922 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
924 while (skb_queue_len(&l
->transmq
) < l
->window
) {
925 skb
= skb_peek(&l
->backlogq
);
928 _skb
= skb_clone(skb
, GFP_ATOMIC
);
931 __skb_dequeue(&l
->backlogq
);
933 l
->backlog
[msg_importance(hdr
)].len
--;
934 __skb_queue_tail(&l
->transmq
, skb
);
935 __skb_queue_tail(xmitq
, _skb
);
936 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
937 msg_set_seqno(hdr
, seqno
);
938 msg_set_ack(hdr
, ack
);
939 msg_set_bcast_ack(hdr
, bc_ack
);
946 static void link_retransmit_failure(struct tipc_link
*l
, struct sk_buff
*skb
)
948 struct tipc_msg
*hdr
= buf_msg(skb
);
950 pr_warn("Retransmission failure on link <%s>\n", l
->name
);
951 link_print(l
, "Resetting link ");
952 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
953 msg_user(hdr
), msg_type(hdr
), msg_size(hdr
), msg_errcode(hdr
));
954 pr_info("sqno %u, prev: %x, src: %x\n",
955 msg_seqno(hdr
), msg_prevnode(hdr
), msg_orignode(hdr
));
958 int tipc_link_retrans(struct tipc_link
*l
, u16 from
, u16 to
,
959 struct sk_buff_head
*xmitq
)
961 struct sk_buff
*_skb
, *skb
= skb_peek(&l
->transmq
);
962 struct tipc_msg
*hdr
;
963 u16 ack
= l
->rcv_nxt
- 1;
964 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
969 /* Detect repeated retransmit failures on same packet */
970 if (likely(l
->last_retransm
!= buf_seqno(skb
))) {
971 l
->last_retransm
= buf_seqno(skb
);
973 } else if (++l
->stale_count
> 100) {
974 link_retransmit_failure(l
, skb
);
975 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
978 /* Move forward to where retransmission should start */
979 skb_queue_walk(&l
->transmq
, skb
) {
980 if (!less(buf_seqno(skb
), from
))
984 skb_queue_walk_from(&l
->transmq
, skb
) {
985 if (more(buf_seqno(skb
), to
))
988 _skb
= __pskb_copy(skb
, MIN_H_SIZE
, GFP_ATOMIC
);
992 msg_set_ack(hdr
, ack
);
993 msg_set_bcast_ack(hdr
, bc_ack
);
994 _skb
->priority
= TC_PRIO_CONTROL
;
995 __skb_queue_tail(xmitq
, _skb
);
996 l
->stats
.retransmitted
++;
1001 /* tipc_data_input - deliver data and name distr msgs to upper layer
1003 * Consumes buffer if message is of right type
1004 * Node lock must be held
1006 static bool tipc_data_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1007 struct sk_buff_head
*inputq
)
1009 switch (msg_user(buf_msg(skb
))) {
1010 case TIPC_LOW_IMPORTANCE
:
1011 case TIPC_MEDIUM_IMPORTANCE
:
1012 case TIPC_HIGH_IMPORTANCE
:
1013 case TIPC_CRITICAL_IMPORTANCE
:
1015 skb_queue_tail(inputq
, skb
);
1017 case NAME_DISTRIBUTOR
:
1018 l
->bc_rcvlink
->state
= LINK_ESTABLISHED
;
1019 skb_queue_tail(l
->namedq
, skb
);
1022 case TUNNEL_PROTOCOL
:
1023 case MSG_FRAGMENTER
:
1024 case BCAST_PROTOCOL
:
1027 pr_warn("Dropping received illegal msg type\n");
1033 /* tipc_link_input - process packet that has passed link protocol check
1037 static int tipc_link_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1038 struct sk_buff_head
*inputq
)
1040 struct tipc_msg
*hdr
= buf_msg(skb
);
1041 struct sk_buff
**reasm_skb
= &l
->reasm_buf
;
1042 struct sk_buff
*iskb
;
1043 struct sk_buff_head tmpq
;
1044 int usr
= msg_user(hdr
);
1049 if (unlikely(usr
== TUNNEL_PROTOCOL
)) {
1050 if (msg_type(hdr
) == SYNCH_MSG
) {
1051 __skb_queue_purge(&l
->deferdq
);
1054 if (!tipc_msg_extract(skb
, &iskb
, &ipos
))
1059 if (less(msg_seqno(hdr
), l
->drop_point
))
1061 if (tipc_data_input(l
, skb
, inputq
))
1063 usr
= msg_user(hdr
);
1064 reasm_skb
= &l
->failover_reasm_skb
;
1067 if (usr
== MSG_BUNDLER
) {
1068 skb_queue_head_init(&tmpq
);
1069 l
->stats
.recv_bundles
++;
1070 l
->stats
.recv_bundled
+= msg_msgcnt(hdr
);
1071 while (tipc_msg_extract(skb
, &iskb
, &pos
))
1072 tipc_data_input(l
, iskb
, &tmpq
);
1073 tipc_skb_queue_splice_tail(&tmpq
, inputq
);
1075 } else if (usr
== MSG_FRAGMENTER
) {
1076 l
->stats
.recv_fragments
++;
1077 if (tipc_buf_append(reasm_skb
, &skb
)) {
1078 l
->stats
.recv_fragmented
++;
1079 tipc_data_input(l
, skb
, inputq
);
1080 } else if (!*reasm_skb
&& !link_is_bc_rcvlink(l
)) {
1081 pr_warn_ratelimited("Unable to build fragment list\n");
1082 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1085 } else if (usr
== BCAST_PROTOCOL
) {
1086 tipc_bcast_lock(l
->net
);
1087 tipc_link_bc_init_rcv(l
->bc_rcvlink
, hdr
);
1088 tipc_bcast_unlock(l
->net
);
1095 static bool tipc_link_release_pkts(struct tipc_link
*l
, u16 acked
)
1097 bool released
= false;
1098 struct sk_buff
*skb
, *tmp
;
1100 skb_queue_walk_safe(&l
->transmq
, skb
, tmp
) {
1101 if (more(buf_seqno(skb
), acked
))
1103 __skb_unlink(skb
, &l
->transmq
);
1110 /* tipc_link_build_state_msg: prepare link state message for transmission
1112 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1113 * risk of ack storms towards the sender
1115 int tipc_link_build_state_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1120 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1121 if (link_is_bc_rcvlink(l
)) {
1122 if (((l
->rcv_nxt
^ tipc_own_addr(l
->net
)) & 0xf) != 0xf)
1125 return TIPC_LINK_SND_BC_ACK
;
1130 l
->stats
.sent_acks
++;
1131 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, xmitq
);
1135 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1137 void tipc_link_build_reset_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1139 int mtyp
= RESET_MSG
;
1140 struct sk_buff
*skb
;
1142 if (l
->state
== LINK_ESTABLISHING
)
1143 mtyp
= ACTIVATE_MSG
;
1145 tipc_link_build_proto_msg(l
, mtyp
, 0, 0, 0, 0, xmitq
);
1147 /* Inform peer that this endpoint is going down if applicable */
1148 skb
= skb_peek_tail(xmitq
);
1149 if (skb
&& (l
->state
== LINK_RESET
))
1150 msg_set_peer_stopping(buf_msg(skb
), 1);
1153 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1155 static void tipc_link_build_nack_msg(struct tipc_link
*l
,
1156 struct sk_buff_head
*xmitq
)
1158 u32 def_cnt
= ++l
->stats
.deferred_recv
;
1160 if (link_is_bc_rcvlink(l
))
1163 if ((skb_queue_len(&l
->deferdq
) == 1) || !(def_cnt
% TIPC_NACK_INTV
))
1164 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, xmitq
);
1167 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1168 * @l: the link that should handle the message
1170 * @xmitq: queue to place packets to be sent after this call
1172 int tipc_link_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1173 struct sk_buff_head
*xmitq
)
1175 struct sk_buff_head
*defq
= &l
->deferdq
;
1176 struct tipc_msg
*hdr
;
1177 u16 seqno
, rcv_nxt
, win_lim
;
1182 seqno
= msg_seqno(hdr
);
1183 rcv_nxt
= l
->rcv_nxt
;
1184 win_lim
= rcv_nxt
+ TIPC_MAX_LINK_WIN
;
1186 /* Verify and update link state */
1187 if (unlikely(msg_user(hdr
) == LINK_PROTOCOL
))
1188 return tipc_link_proto_rcv(l
, skb
, xmitq
);
1190 if (unlikely(!link_is_up(l
))) {
1191 if (l
->state
== LINK_ESTABLISHING
)
1192 rc
= TIPC_LINK_UP_EVT
;
1196 /* Don't send probe at next timeout expiration */
1197 l
->silent_intv_cnt
= 0;
1199 /* Drop if outside receive window */
1200 if (unlikely(less(seqno
, rcv_nxt
) || more(seqno
, win_lim
))) {
1201 l
->stats
.duplicates
++;
1205 /* Forward queues and wake up waiting users */
1206 if (likely(tipc_link_release_pkts(l
, msg_ack(hdr
)))) {
1207 tipc_link_advance_backlog(l
, xmitq
);
1208 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
1209 link_prepare_wakeup(l
);
1212 /* Defer delivery if sequence gap */
1213 if (unlikely(seqno
!= rcv_nxt
)) {
1214 __tipc_skb_queue_sorted(defq
, seqno
, skb
);
1215 tipc_link_build_nack_msg(l
, xmitq
);
1219 /* Deliver packet */
1221 l
->stats
.recv_info
++;
1222 if (!tipc_data_input(l
, skb
, l
->inputq
))
1223 rc
|= tipc_link_input(l
, skb
, l
->inputq
);
1224 if (unlikely(++l
->rcv_unacked
>= TIPC_MIN_LINK_WIN
))
1225 rc
|= tipc_link_build_state_msg(l
, xmitq
);
1226 if (unlikely(rc
& ~TIPC_LINK_SND_BC_ACK
))
1228 } while ((skb
= __skb_dequeue(defq
)));
1236 static void tipc_link_build_proto_msg(struct tipc_link
*l
, int mtyp
, bool probe
,
1237 u16 rcvgap
, int tolerance
, int priority
,
1238 struct sk_buff_head
*xmitq
)
1240 struct sk_buff
*skb
;
1241 struct tipc_msg
*hdr
;
1242 struct sk_buff_head
*dfq
= &l
->deferdq
;
1243 bool node_up
= link_is_up(l
->bc_rcvlink
);
1245 /* Don't send protocol message during reset or link failover */
1246 if (tipc_link_is_blocked(l
))
1249 if (!tipc_link_is_up(l
) && (mtyp
== STATE_MSG
))
1252 if (!skb_queue_empty(dfq
))
1253 rcvgap
= buf_seqno(skb_peek(dfq
)) - l
->rcv_nxt
;
1255 skb
= tipc_msg_create(LINK_PROTOCOL
, mtyp
, INT_H_SIZE
,
1256 TIPC_MAX_IF_NAME
, l
->addr
,
1257 tipc_own_addr(l
->net
), 0, 0, 0);
1262 msg_set_session(hdr
, l
->session
);
1263 msg_set_bearer_id(hdr
, l
->bearer_id
);
1264 msg_set_net_plane(hdr
, l
->net_plane
);
1265 msg_set_next_sent(hdr
, l
->snd_nxt
);
1266 msg_set_ack(hdr
, l
->rcv_nxt
- 1);
1267 msg_set_bcast_ack(hdr
, l
->bc_rcvlink
->rcv_nxt
- 1);
1268 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
1269 msg_set_link_tolerance(hdr
, tolerance
);
1270 msg_set_linkprio(hdr
, priority
);
1271 msg_set_redundant_link(hdr
, node_up
);
1272 msg_set_seq_gap(hdr
, 0);
1273 msg_set_seqno(hdr
, l
->snd_nxt
+ U16_MAX
/ 2);
1275 if (mtyp
== STATE_MSG
) {
1276 msg_set_seq_gap(hdr
, rcvgap
);
1277 msg_set_size(hdr
, INT_H_SIZE
);
1278 msg_set_probe(hdr
, probe
);
1279 l
->stats
.sent_states
++;
1282 /* RESET_MSG or ACTIVATE_MSG */
1283 msg_set_max_pkt(hdr
, l
->advertised_mtu
);
1284 strcpy(msg_data(hdr
), l
->if_name
);
1287 l
->stats
.sent_probes
++;
1289 l
->stats
.sent_nacks
++;
1290 skb
->priority
= TC_PRIO_CONTROL
;
1291 __skb_queue_tail(xmitq
, skb
);
1294 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1295 * with contents of the link's transmit and backlog queues.
1297 void tipc_link_tnl_prepare(struct tipc_link
*l
, struct tipc_link
*tnl
,
1298 int mtyp
, struct sk_buff_head
*xmitq
)
1300 struct sk_buff
*skb
, *tnlskb
;
1301 struct tipc_msg
*hdr
, tnlhdr
;
1302 struct sk_buff_head
*queue
= &l
->transmq
;
1303 struct sk_buff_head tmpxq
, tnlq
;
1304 u16 pktlen
, pktcnt
, seqno
= l
->snd_nxt
;
1309 skb_queue_head_init(&tnlq
);
1310 skb_queue_head_init(&tmpxq
);
1312 /* At least one packet required for safe algorithm => add dummy */
1313 skb
= tipc_msg_create(TIPC_LOW_IMPORTANCE
, TIPC_DIRECT_MSG
,
1314 BASIC_H_SIZE
, 0, l
->addr
, tipc_own_addr(l
->net
),
1315 0, 0, TIPC_ERR_NO_PORT
);
1317 pr_warn("%sunable to create tunnel packet\n", link_co_err
);
1320 skb_queue_tail(&tnlq
, skb
);
1321 tipc_link_xmit(l
, &tnlq
, &tmpxq
);
1322 __skb_queue_purge(&tmpxq
);
1324 /* Initialize reusable tunnel packet header */
1325 tipc_msg_init(tipc_own_addr(l
->net
), &tnlhdr
, TUNNEL_PROTOCOL
,
1326 mtyp
, INT_H_SIZE
, l
->addr
);
1327 pktcnt
= skb_queue_len(&l
->transmq
) + skb_queue_len(&l
->backlogq
);
1328 msg_set_msgcnt(&tnlhdr
, pktcnt
);
1329 msg_set_bearer_id(&tnlhdr
, l
->peer_bearer_id
);
1331 /* Wrap each packet into a tunnel packet */
1332 skb_queue_walk(queue
, skb
) {
1334 if (queue
== &l
->backlogq
)
1335 msg_set_seqno(hdr
, seqno
++);
1336 pktlen
= msg_size(hdr
);
1337 msg_set_size(&tnlhdr
, pktlen
+ INT_H_SIZE
);
1338 tnlskb
= tipc_buf_acquire(pktlen
+ INT_H_SIZE
);
1340 pr_warn("%sunable to send packet\n", link_co_err
);
1343 skb_copy_to_linear_data(tnlskb
, &tnlhdr
, INT_H_SIZE
);
1344 skb_copy_to_linear_data_offset(tnlskb
, INT_H_SIZE
, hdr
, pktlen
);
1345 __skb_queue_tail(&tnlq
, tnlskb
);
1347 if (queue
!= &l
->backlogq
) {
1348 queue
= &l
->backlogq
;
1352 tipc_link_xmit(tnl
, &tnlq
, xmitq
);
1354 if (mtyp
== FAILOVER_MSG
) {
1355 tnl
->drop_point
= l
->rcv_nxt
;
1356 tnl
->failover_reasm_skb
= l
->reasm_buf
;
1357 l
->reasm_buf
= NULL
;
1361 /* tipc_link_proto_rcv(): receive link level protocol message :
1362 * Note that network plane id propagates through the network, and may
1363 * change at any time. The node with lowest numerical id determines
1366 static int tipc_link_proto_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1367 struct sk_buff_head
*xmitq
)
1369 struct tipc_msg
*hdr
= buf_msg(skb
);
1371 u16 ack
= msg_ack(hdr
);
1372 u16 gap
= msg_seq_gap(hdr
);
1373 u16 peers_snd_nxt
= msg_next_sent(hdr
);
1374 u16 peers_tol
= msg_link_tolerance(hdr
);
1375 u16 peers_prio
= msg_linkprio(hdr
);
1376 u16 rcv_nxt
= l
->rcv_nxt
;
1377 int mtyp
= msg_type(hdr
);
1381 if (tipc_link_is_blocked(l
) || !xmitq
)
1384 if (tipc_own_addr(l
->net
) > msg_prevnode(hdr
))
1385 l
->net_plane
= msg_net_plane(hdr
);
1390 /* Ignore duplicate RESET with old session number */
1391 if ((less_eq(msg_session(hdr
), l
->peer_session
)) &&
1392 (l
->peer_session
!= ANY_SESSION
))
1400 /* Complete own link name with peer's interface name */
1401 if_name
= strrchr(l
->name
, ':') + 1;
1402 if (sizeof(l
->name
) - (if_name
- l
->name
) <= TIPC_MAX_IF_NAME
)
1404 if (msg_data_sz(hdr
) < TIPC_MAX_IF_NAME
)
1406 strncpy(if_name
, msg_data(hdr
), TIPC_MAX_IF_NAME
);
1408 /* Update own tolerance if peer indicates a non-zero value */
1409 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
))
1410 l
->tolerance
= peers_tol
;
1412 /* Update own priority if peer's priority is higher */
1413 if (in_range(peers_prio
, l
->priority
+ 1, TIPC_MAX_LINK_PRI
))
1414 l
->priority
= peers_prio
;
1416 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1417 if (msg_peer_stopping(hdr
))
1418 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1419 else if ((mtyp
== RESET_MSG
) || !link_is_up(l
))
1420 rc
= tipc_link_fsm_evt(l
, LINK_PEER_RESET_EVT
);
1422 /* ACTIVATE_MSG takes up link if it was already locally reset */
1423 if ((mtyp
== ACTIVATE_MSG
) && (l
->state
== LINK_ESTABLISHING
))
1424 rc
= TIPC_LINK_UP_EVT
;
1426 l
->peer_session
= msg_session(hdr
);
1427 l
->peer_bearer_id
= msg_bearer_id(hdr
);
1428 if (l
->mtu
> msg_max_pkt(hdr
))
1429 l
->mtu
= msg_max_pkt(hdr
);
1434 /* Update own tolerance if peer indicates a non-zero value */
1435 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
))
1436 l
->tolerance
= peers_tol
;
1438 if (peers_prio
&& in_range(peers_prio
, TIPC_MIN_LINK_PRI
,
1439 TIPC_MAX_LINK_PRI
)) {
1440 l
->priority
= peers_prio
;
1441 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1444 l
->silent_intv_cnt
= 0;
1445 l
->stats
.recv_states
++;
1447 l
->stats
.recv_probes
++;
1449 if (!link_is_up(l
)) {
1450 if (l
->state
== LINK_ESTABLISHING
)
1451 rc
= TIPC_LINK_UP_EVT
;
1455 /* Send NACK if peer has sent pkts we haven't received yet */
1456 if (more(peers_snd_nxt
, rcv_nxt
) && !tipc_link_is_synching(l
))
1457 rcvgap
= peers_snd_nxt
- l
->rcv_nxt
;
1458 if (rcvgap
|| (msg_probe(hdr
)))
1459 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, rcvgap
,
1461 tipc_link_release_pkts(l
, ack
);
1463 /* If NACK, retransmit will now start at right position */
1465 rc
= tipc_link_retrans(l
, ack
+ 1, ack
+ gap
, xmitq
);
1466 l
->stats
.recv_nacks
++;
1469 tipc_link_advance_backlog(l
, xmitq
);
1470 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
1471 link_prepare_wakeup(l
);
1478 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1480 static bool tipc_link_build_bc_proto_msg(struct tipc_link
*l
, bool bcast
,
1482 struct sk_buff_head
*xmitq
)
1484 struct sk_buff
*skb
;
1485 struct tipc_msg
*hdr
;
1486 struct sk_buff
*dfrd_skb
= skb_peek(&l
->deferdq
);
1487 u16 ack
= l
->rcv_nxt
- 1;
1488 u16 gap_to
= peers_snd_nxt
- 1;
1490 skb
= tipc_msg_create(BCAST_PROTOCOL
, STATE_MSG
, INT_H_SIZE
,
1491 0, l
->addr
, tipc_own_addr(l
->net
), 0, 0, 0);
1495 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
1496 msg_set_bcast_ack(hdr
, ack
);
1497 msg_set_bcgap_after(hdr
, ack
);
1499 gap_to
= buf_seqno(dfrd_skb
) - 1;
1500 msg_set_bcgap_to(hdr
, gap_to
);
1501 msg_set_non_seq(hdr
, bcast
);
1502 __skb_queue_tail(xmitq
, skb
);
1506 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1508 * Give a newly added peer node the sequence number where it should
1509 * start receiving and acking broadcast packets.
1511 static void tipc_link_build_bc_init_msg(struct tipc_link
*l
,
1512 struct sk_buff_head
*xmitq
)
1514 struct sk_buff_head list
;
1516 __skb_queue_head_init(&list
);
1517 if (!tipc_link_build_bc_proto_msg(l
->bc_rcvlink
, false, 0, &list
))
1519 tipc_link_xmit(l
, &list
, xmitq
);
1522 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1524 void tipc_link_bc_init_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
)
1526 int mtyp
= msg_type(hdr
);
1527 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
1532 if (msg_user(hdr
) == BCAST_PROTOCOL
) {
1533 l
->rcv_nxt
= peers_snd_nxt
;
1534 l
->state
= LINK_ESTABLISHED
;
1538 if (l
->peer_caps
& TIPC_BCAST_SYNCH
)
1541 if (msg_peer_node_is_up(hdr
))
1544 /* Compatibility: accept older, less safe initial synch data */
1545 if ((mtyp
== RESET_MSG
) || (mtyp
== ACTIVATE_MSG
))
1546 l
->rcv_nxt
= peers_snd_nxt
;
1549 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1551 void tipc_link_bc_sync_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
,
1552 struct sk_buff_head
*xmitq
)
1554 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
1559 if (!msg_peer_node_is_up(hdr
))
1562 l
->bc_peer_is_up
= true;
1564 /* Ignore if peers_snd_nxt goes beyond receive window */
1565 if (more(peers_snd_nxt
, l
->rcv_nxt
+ l
->window
))
1568 if (!more(peers_snd_nxt
, l
->rcv_nxt
)) {
1569 l
->nack_state
= BC_NACK_SND_CONDITIONAL
;
1573 /* Don't NACK if one was recently sent or peeked */
1574 if (l
->nack_state
== BC_NACK_SND_SUPPRESS
) {
1575 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
1579 /* Conditionally delay NACK sending until next synch rcv */
1580 if (l
->nack_state
== BC_NACK_SND_CONDITIONAL
) {
1581 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
1582 if ((peers_snd_nxt
- l
->rcv_nxt
) < TIPC_MIN_LINK_WIN
)
1586 /* Send NACK now but suppress next one */
1587 tipc_link_build_bc_proto_msg(l
, true, peers_snd_nxt
, xmitq
);
1588 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
1591 void tipc_link_bc_ack_rcv(struct tipc_link
*l
, u16 acked
,
1592 struct sk_buff_head
*xmitq
)
1594 struct sk_buff
*skb
, *tmp
;
1595 struct tipc_link
*snd_l
= l
->bc_sndlink
;
1597 if (!link_is_up(l
) || !l
->bc_peer_is_up
)
1600 if (!more(acked
, l
->acked
))
1603 /* Skip over packets peer has already acked */
1604 skb_queue_walk(&snd_l
->transmq
, skb
) {
1605 if (more(buf_seqno(skb
), l
->acked
))
1609 /* Update/release the packets peer is acking now */
1610 skb_queue_walk_from_safe(&snd_l
->transmq
, skb
, tmp
) {
1611 if (more(buf_seqno(skb
), acked
))
1613 if (!--TIPC_SKB_CB(skb
)->ackers
) {
1614 __skb_unlink(skb
, &snd_l
->transmq
);
1619 tipc_link_advance_backlog(snd_l
, xmitq
);
1620 if (unlikely(!skb_queue_empty(&snd_l
->wakeupq
)))
1621 link_prepare_wakeup(snd_l
);
1624 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1626 int tipc_link_bc_nack_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1627 struct sk_buff_head
*xmitq
)
1629 struct tipc_msg
*hdr
= buf_msg(skb
);
1630 u32 dnode
= msg_destnode(hdr
);
1631 int mtyp
= msg_type(hdr
);
1632 u16 acked
= msg_bcast_ack(hdr
);
1633 u16 from
= acked
+ 1;
1634 u16 to
= msg_bcgap_to(hdr
);
1635 u16 peers_snd_nxt
= to
+ 1;
1640 if (!tipc_link_is_up(l
) || !l
->bc_peer_is_up
)
1643 if (mtyp
!= STATE_MSG
)
1646 if (dnode
== tipc_own_addr(l
->net
)) {
1647 tipc_link_bc_ack_rcv(l
, acked
, xmitq
);
1648 rc
= tipc_link_retrans(l
->bc_sndlink
, from
, to
, xmitq
);
1649 l
->stats
.recv_nacks
++;
1653 /* Msg for other node => suppress own NACK at next sync if applicable */
1654 if (more(peers_snd_nxt
, l
->rcv_nxt
) && !less(l
->rcv_nxt
, from
))
1655 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
1660 void tipc_link_set_queue_limits(struct tipc_link
*l
, u32 win
)
1662 int max_bulk
= TIPC_MAX_PUBLICATIONS
/ (l
->mtu
/ ITEM_SIZE
);
1665 l
->backlog
[TIPC_LOW_IMPORTANCE
].limit
= win
/ 2;
1666 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].limit
= win
;
1667 l
->backlog
[TIPC_HIGH_IMPORTANCE
].limit
= win
/ 2 * 3;
1668 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].limit
= win
* 2;
1669 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].limit
= max_bulk
;
1673 * link_reset_stats - reset link statistics
1674 * @l: pointer to link
1676 void tipc_link_reset_stats(struct tipc_link
*l
)
1678 memset(&l
->stats
, 0, sizeof(l
->stats
));
1679 if (!link_is_bc_sndlink(l
)) {
1680 l
->stats
.sent_info
= l
->snd_nxt
;
1681 l
->stats
.recv_info
= l
->rcv_nxt
;
1685 static void link_print(struct tipc_link
*l
, const char *str
)
1687 struct sk_buff
*hskb
= skb_peek(&l
->transmq
);
1688 u16 head
= hskb
? msg_seqno(buf_msg(hskb
)) : l
->snd_nxt
- 1;
1689 u16 tail
= l
->snd_nxt
- 1;
1691 pr_info("%s Link <%s> state %x\n", str
, l
->name
, l
->state
);
1692 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1693 skb_queue_len(&l
->transmq
), head
, tail
,
1694 skb_queue_len(&l
->backlogq
), l
->snd_nxt
, l
->rcv_nxt
);
1697 /* Parse and validate nested (link) properties valid for media, bearer and link
1699 int tipc_nl_parse_link_prop(struct nlattr
*prop
, struct nlattr
*props
[])
1703 err
= nla_parse_nested(props
, TIPC_NLA_PROP_MAX
, prop
,
1704 tipc_nl_prop_policy
);
1708 if (props
[TIPC_NLA_PROP_PRIO
]) {
1711 prio
= nla_get_u32(props
[TIPC_NLA_PROP_PRIO
]);
1712 if (prio
> TIPC_MAX_LINK_PRI
)
1716 if (props
[TIPC_NLA_PROP_TOL
]) {
1719 tol
= nla_get_u32(props
[TIPC_NLA_PROP_TOL
]);
1720 if ((tol
< TIPC_MIN_LINK_TOL
) || (tol
> TIPC_MAX_LINK_TOL
))
1724 if (props
[TIPC_NLA_PROP_WIN
]) {
1727 win
= nla_get_u32(props
[TIPC_NLA_PROP_WIN
]);
1728 if ((win
< TIPC_MIN_LINK_WIN
) || (win
> TIPC_MAX_LINK_WIN
))
1735 static int __tipc_nl_add_stats(struct sk_buff
*skb
, struct tipc_stats
*s
)
1738 struct nlattr
*stats
;
1745 struct nla_map map
[] = {
1746 {TIPC_NLA_STATS_RX_INFO
, s
->recv_info
},
1747 {TIPC_NLA_STATS_RX_FRAGMENTS
, s
->recv_fragments
},
1748 {TIPC_NLA_STATS_RX_FRAGMENTED
, s
->recv_fragmented
},
1749 {TIPC_NLA_STATS_RX_BUNDLES
, s
->recv_bundles
},
1750 {TIPC_NLA_STATS_RX_BUNDLED
, s
->recv_bundled
},
1751 {TIPC_NLA_STATS_TX_INFO
, s
->sent_info
},
1752 {TIPC_NLA_STATS_TX_FRAGMENTS
, s
->sent_fragments
},
1753 {TIPC_NLA_STATS_TX_FRAGMENTED
, s
->sent_fragmented
},
1754 {TIPC_NLA_STATS_TX_BUNDLES
, s
->sent_bundles
},
1755 {TIPC_NLA_STATS_TX_BUNDLED
, s
->sent_bundled
},
1756 {TIPC_NLA_STATS_MSG_PROF_TOT
, (s
->msg_length_counts
) ?
1757 s
->msg_length_counts
: 1},
1758 {TIPC_NLA_STATS_MSG_LEN_CNT
, s
->msg_length_counts
},
1759 {TIPC_NLA_STATS_MSG_LEN_TOT
, s
->msg_lengths_total
},
1760 {TIPC_NLA_STATS_MSG_LEN_P0
, s
->msg_length_profile
[0]},
1761 {TIPC_NLA_STATS_MSG_LEN_P1
, s
->msg_length_profile
[1]},
1762 {TIPC_NLA_STATS_MSG_LEN_P2
, s
->msg_length_profile
[2]},
1763 {TIPC_NLA_STATS_MSG_LEN_P3
, s
->msg_length_profile
[3]},
1764 {TIPC_NLA_STATS_MSG_LEN_P4
, s
->msg_length_profile
[4]},
1765 {TIPC_NLA_STATS_MSG_LEN_P5
, s
->msg_length_profile
[5]},
1766 {TIPC_NLA_STATS_MSG_LEN_P6
, s
->msg_length_profile
[6]},
1767 {TIPC_NLA_STATS_RX_STATES
, s
->recv_states
},
1768 {TIPC_NLA_STATS_RX_PROBES
, s
->recv_probes
},
1769 {TIPC_NLA_STATS_RX_NACKS
, s
->recv_nacks
},
1770 {TIPC_NLA_STATS_RX_DEFERRED
, s
->deferred_recv
},
1771 {TIPC_NLA_STATS_TX_STATES
, s
->sent_states
},
1772 {TIPC_NLA_STATS_TX_PROBES
, s
->sent_probes
},
1773 {TIPC_NLA_STATS_TX_NACKS
, s
->sent_nacks
},
1774 {TIPC_NLA_STATS_TX_ACKS
, s
->sent_acks
},
1775 {TIPC_NLA_STATS_RETRANSMITTED
, s
->retransmitted
},
1776 {TIPC_NLA_STATS_DUPLICATES
, s
->duplicates
},
1777 {TIPC_NLA_STATS_LINK_CONGS
, s
->link_congs
},
1778 {TIPC_NLA_STATS_MAX_QUEUE
, s
->max_queue_sz
},
1779 {TIPC_NLA_STATS_AVG_QUEUE
, s
->queue_sz_counts
?
1780 (s
->accu_queue_sz
/ s
->queue_sz_counts
) : 0}
1783 stats
= nla_nest_start(skb
, TIPC_NLA_LINK_STATS
);
1787 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
1788 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
1791 nla_nest_end(skb
, stats
);
1795 nla_nest_cancel(skb
, stats
);
1800 /* Caller should hold appropriate locks to protect the link */
1801 int __tipc_nl_add_link(struct net
*net
, struct tipc_nl_msg
*msg
,
1802 struct tipc_link
*link
, int nlflags
)
1806 struct nlattr
*attrs
;
1807 struct nlattr
*prop
;
1808 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
1810 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
1811 nlflags
, TIPC_NL_LINK_GET
);
1815 attrs
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK
);
1819 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, link
->name
))
1821 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_DEST
,
1822 tipc_cluster_mask(tn
->own_addr
)))
1824 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_MTU
, link
->mtu
))
1826 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, link
->rcv_nxt
))
1828 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, link
->snd_nxt
))
1831 if (tipc_link_is_up(link
))
1832 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
1835 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_ACTIVE
))
1838 prop
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK_PROP
);
1841 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
1843 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_TOL
, link
->tolerance
))
1845 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
,
1848 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
1850 nla_nest_end(msg
->skb
, prop
);
1852 err
= __tipc_nl_add_stats(msg
->skb
, &link
->stats
);
1856 nla_nest_end(msg
->skb
, attrs
);
1857 genlmsg_end(msg
->skb
, hdr
);
1862 nla_nest_cancel(msg
->skb
, prop
);
1864 nla_nest_cancel(msg
->skb
, attrs
);
1866 genlmsg_cancel(msg
->skb
, hdr
);
1871 static int __tipc_nl_add_bc_link_stat(struct sk_buff
*skb
,
1872 struct tipc_stats
*stats
)
1875 struct nlattr
*nest
;
1882 struct nla_map map
[] = {
1883 {TIPC_NLA_STATS_RX_INFO
, stats
->recv_info
},
1884 {TIPC_NLA_STATS_RX_FRAGMENTS
, stats
->recv_fragments
},
1885 {TIPC_NLA_STATS_RX_FRAGMENTED
, stats
->recv_fragmented
},
1886 {TIPC_NLA_STATS_RX_BUNDLES
, stats
->recv_bundles
},
1887 {TIPC_NLA_STATS_RX_BUNDLED
, stats
->recv_bundled
},
1888 {TIPC_NLA_STATS_TX_INFO
, stats
->sent_info
},
1889 {TIPC_NLA_STATS_TX_FRAGMENTS
, stats
->sent_fragments
},
1890 {TIPC_NLA_STATS_TX_FRAGMENTED
, stats
->sent_fragmented
},
1891 {TIPC_NLA_STATS_TX_BUNDLES
, stats
->sent_bundles
},
1892 {TIPC_NLA_STATS_TX_BUNDLED
, stats
->sent_bundled
},
1893 {TIPC_NLA_STATS_RX_NACKS
, stats
->recv_nacks
},
1894 {TIPC_NLA_STATS_RX_DEFERRED
, stats
->deferred_recv
},
1895 {TIPC_NLA_STATS_TX_NACKS
, stats
->sent_nacks
},
1896 {TIPC_NLA_STATS_TX_ACKS
, stats
->sent_acks
},
1897 {TIPC_NLA_STATS_RETRANSMITTED
, stats
->retransmitted
},
1898 {TIPC_NLA_STATS_DUPLICATES
, stats
->duplicates
},
1899 {TIPC_NLA_STATS_LINK_CONGS
, stats
->link_congs
},
1900 {TIPC_NLA_STATS_MAX_QUEUE
, stats
->max_queue_sz
},
1901 {TIPC_NLA_STATS_AVG_QUEUE
, stats
->queue_sz_counts
?
1902 (stats
->accu_queue_sz
/ stats
->queue_sz_counts
) : 0}
1905 nest
= nla_nest_start(skb
, TIPC_NLA_LINK_STATS
);
1909 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
1910 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
1913 nla_nest_end(skb
, nest
);
1917 nla_nest_cancel(skb
, nest
);
1922 int tipc_nl_add_bc_link(struct net
*net
, struct tipc_nl_msg
*msg
)
1926 struct nlattr
*attrs
;
1927 struct nlattr
*prop
;
1928 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
1929 struct tipc_link
*bcl
= tn
->bcl
;
1934 tipc_bcast_lock(net
);
1936 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
1937 NLM_F_MULTI
, TIPC_NL_LINK_GET
);
1939 tipc_bcast_unlock(net
);
1943 attrs
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK
);
1947 /* The broadcast link is always up */
1948 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
1951 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_BROADCAST
))
1953 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, bcl
->name
))
1955 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, bcl
->rcv_nxt
))
1957 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, bcl
->snd_nxt
))
1960 prop
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK_PROP
);
1963 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
, bcl
->window
))
1965 nla_nest_end(msg
->skb
, prop
);
1967 err
= __tipc_nl_add_bc_link_stat(msg
->skb
, &bcl
->stats
);
1971 tipc_bcast_unlock(net
);
1972 nla_nest_end(msg
->skb
, attrs
);
1973 genlmsg_end(msg
->skb
, hdr
);
1978 nla_nest_cancel(msg
->skb
, prop
);
1980 nla_nest_cancel(msg
->skb
, attrs
);
1982 tipc_bcast_unlock(net
);
1983 genlmsg_cancel(msg
->skb
, hdr
);
1988 void tipc_link_set_tolerance(struct tipc_link
*l
, u32 tol
,
1989 struct sk_buff_head
*xmitq
)
1992 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, tol
, 0, xmitq
);
1995 void tipc_link_set_prio(struct tipc_link
*l
, u32 prio
,
1996 struct sk_buff_head
*xmitq
)
1999 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, prio
, xmitq
);
2002 void tipc_link_set_abort_limit(struct tipc_link
*l
, u32 limit
)
2004 l
->abort_limit
= limit
;