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
)
711 u16 bc_snt
= l
->bc_sndlink
->snd_nxt
- 1;
712 u16 bc_acked
= l
->bc_rcvlink
->acked
;
714 link_profile_stats(l
);
717 case LINK_ESTABLISHED
:
719 if (l
->silent_intv_cnt
> l
->abort_limit
)
720 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
722 state
= bc_acked
!= bc_snt
;
723 probe
= l
->silent_intv_cnt
;
724 l
->silent_intv_cnt
++;
727 setup
= l
->rst_cnt
++ <= 4;
728 setup
|= !(l
->rst_cnt
% 16);
731 case LINK_ESTABLISHING
:
735 case LINK_PEER_RESET
:
737 case LINK_FAILINGOVER
:
743 if (state
|| probe
|| setup
)
744 tipc_link_build_proto_msg(l
, mtyp
, probe
, 0, 0, 0, xmitq
);
750 * link_schedule_user - schedule a message sender for wakeup after congestion
751 * @link: congested link
752 * @list: message that was attempted sent
753 * Create pseudo msg to send back to user when congestion abates
754 * Does not consume buffer list
756 static int link_schedule_user(struct tipc_link
*link
, struct sk_buff_head
*list
)
758 struct tipc_msg
*msg
= buf_msg(skb_peek(list
));
759 int imp
= msg_importance(msg
);
760 u32 oport
= msg_origport(msg
);
761 u32 addr
= tipc_own_addr(link
->net
);
764 /* This really cannot happen... */
765 if (unlikely(imp
> TIPC_CRITICAL_IMPORTANCE
)) {
766 pr_warn("%s<%s>, send queue full", link_rst_msg
, link
->name
);
769 /* Non-blocking sender: */
770 if (TIPC_SKB_CB(skb_peek(list
))->wakeup_pending
)
773 /* Create and schedule wakeup pseudo message */
774 skb
= tipc_msg_create(SOCK_WAKEUP
, 0, INT_H_SIZE
, 0,
775 addr
, addr
, oport
, 0, 0);
778 TIPC_SKB_CB(skb
)->chain_sz
= skb_queue_len(list
);
779 TIPC_SKB_CB(skb
)->chain_imp
= imp
;
780 skb_queue_tail(&link
->wakeupq
, skb
);
781 link
->stats
.link_congs
++;
786 * link_prepare_wakeup - prepare users for wakeup after congestion
787 * @link: congested link
788 * Move a number of waiting users, as permitted by available space in
789 * the send queue, from link wait queue to node wait queue for wakeup
791 void link_prepare_wakeup(struct tipc_link
*l
)
793 int pnd
[TIPC_SYSTEM_IMPORTANCE
+ 1] = {0,};
795 struct sk_buff
*skb
, *tmp
;
797 skb_queue_walk_safe(&l
->wakeupq
, skb
, tmp
) {
798 imp
= TIPC_SKB_CB(skb
)->chain_imp
;
799 lim
= l
->window
+ l
->backlog
[imp
].limit
;
800 pnd
[imp
] += TIPC_SKB_CB(skb
)->chain_sz
;
801 if ((pnd
[imp
] + l
->backlog
[imp
].len
) >= lim
)
803 skb_unlink(skb
, &l
->wakeupq
);
804 skb_queue_tail(l
->inputq
, skb
);
808 void tipc_link_reset(struct tipc_link
*l
)
810 l
->peer_session
= ANY_SESSION
;
812 l
->mtu
= l
->advertised_mtu
;
813 __skb_queue_purge(&l
->transmq
);
814 __skb_queue_purge(&l
->deferdq
);
815 skb_queue_splice_init(&l
->wakeupq
, l
->inputq
);
816 __skb_queue_purge(&l
->backlogq
);
817 l
->backlog
[TIPC_LOW_IMPORTANCE
].len
= 0;
818 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].len
= 0;
819 l
->backlog
[TIPC_HIGH_IMPORTANCE
].len
= 0;
820 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].len
= 0;
821 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].len
= 0;
822 kfree_skb(l
->reasm_buf
);
823 kfree_skb(l
->failover_reasm_skb
);
825 l
->failover_reasm_skb
= NULL
;
830 l
->silent_intv_cnt
= 0;
832 l
->stats
.recv_info
= 0;
834 l
->bc_peer_is_up
= false;
835 tipc_link_reset_stats(l
);
839 * tipc_link_xmit(): enqueue buffer list according to queue situation
841 * @list: chain of buffers containing message
842 * @xmitq: returned list of packets to be sent by caller
844 * Consumes the buffer chain, except when returning -ELINKCONG,
845 * since the caller then may want to make more send attempts.
846 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
847 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
849 int tipc_link_xmit(struct tipc_link
*l
, struct sk_buff_head
*list
,
850 struct sk_buff_head
*xmitq
)
852 struct tipc_msg
*hdr
= buf_msg(skb_peek(list
));
853 unsigned int maxwin
= l
->window
;
854 unsigned int i
, imp
= msg_importance(hdr
);
855 unsigned int mtu
= l
->mtu
;
856 u16 ack
= l
->rcv_nxt
- 1;
857 u16 seqno
= l
->snd_nxt
;
858 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
859 struct sk_buff_head
*transmq
= &l
->transmq
;
860 struct sk_buff_head
*backlogq
= &l
->backlogq
;
861 struct sk_buff
*skb
, *_skb
, *bskb
;
863 /* Match msg importance against this and all higher backlog limits: */
864 for (i
= imp
; i
<= TIPC_SYSTEM_IMPORTANCE
; i
++) {
865 if (unlikely(l
->backlog
[i
].len
>= l
->backlog
[i
].limit
))
866 return link_schedule_user(l
, list
);
868 if (unlikely(msg_size(hdr
) > mtu
)) {
869 skb_queue_purge(list
);
873 /* Prepare each packet for sending, and add to relevant queue: */
874 while (skb_queue_len(list
)) {
875 skb
= skb_peek(list
);
877 msg_set_seqno(hdr
, seqno
);
878 msg_set_ack(hdr
, ack
);
879 msg_set_bcast_ack(hdr
, bc_ack
);
881 if (likely(skb_queue_len(transmq
) < maxwin
)) {
882 _skb
= skb_clone(skb
, GFP_ATOMIC
);
884 skb_queue_purge(list
);
888 __skb_queue_tail(transmq
, skb
);
889 __skb_queue_tail(xmitq
, _skb
);
890 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
895 if (tipc_msg_bundle(skb_peek_tail(backlogq
), hdr
, mtu
)) {
896 kfree_skb(__skb_dequeue(list
));
897 l
->stats
.sent_bundled
++;
900 if (tipc_msg_make_bundle(&bskb
, hdr
, mtu
, l
->addr
)) {
901 kfree_skb(__skb_dequeue(list
));
902 __skb_queue_tail(backlogq
, bskb
);
903 l
->backlog
[msg_importance(buf_msg(bskb
))].len
++;
904 l
->stats
.sent_bundled
++;
905 l
->stats
.sent_bundles
++;
908 l
->backlog
[imp
].len
+= skb_queue_len(list
);
909 skb_queue_splice_tail_init(list
, backlogq
);
915 void tipc_link_advance_backlog(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
917 struct sk_buff
*skb
, *_skb
;
918 struct tipc_msg
*hdr
;
919 u16 seqno
= l
->snd_nxt
;
920 u16 ack
= l
->rcv_nxt
- 1;
921 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
923 while (skb_queue_len(&l
->transmq
) < l
->window
) {
924 skb
= skb_peek(&l
->backlogq
);
927 _skb
= skb_clone(skb
, GFP_ATOMIC
);
930 __skb_dequeue(&l
->backlogq
);
932 l
->backlog
[msg_importance(hdr
)].len
--;
933 __skb_queue_tail(&l
->transmq
, skb
);
934 __skb_queue_tail(xmitq
, _skb
);
935 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
936 msg_set_seqno(hdr
, seqno
);
937 msg_set_ack(hdr
, ack
);
938 msg_set_bcast_ack(hdr
, bc_ack
);
945 static void link_retransmit_failure(struct tipc_link
*l
, struct sk_buff
*skb
)
947 struct tipc_msg
*hdr
= buf_msg(skb
);
949 pr_warn("Retransmission failure on link <%s>\n", l
->name
);
950 link_print(l
, "Resetting link ");
951 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
952 msg_user(hdr
), msg_type(hdr
), msg_size(hdr
), msg_errcode(hdr
));
953 pr_info("sqno %u, prev: %x, src: %x\n",
954 msg_seqno(hdr
), msg_prevnode(hdr
), msg_orignode(hdr
));
957 int tipc_link_retrans(struct tipc_link
*l
, u16 from
, u16 to
,
958 struct sk_buff_head
*xmitq
)
960 struct sk_buff
*_skb
, *skb
= skb_peek(&l
->transmq
);
961 struct tipc_msg
*hdr
;
962 u16 ack
= l
->rcv_nxt
- 1;
963 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
968 /* Detect repeated retransmit failures on same packet */
969 if (likely(l
->last_retransm
!= buf_seqno(skb
))) {
970 l
->last_retransm
= buf_seqno(skb
);
972 } else if (++l
->stale_count
> 100) {
973 link_retransmit_failure(l
, skb
);
974 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
977 /* Move forward to where retransmission should start */
978 skb_queue_walk(&l
->transmq
, skb
) {
979 if (!less(buf_seqno(skb
), from
))
983 skb_queue_walk_from(&l
->transmq
, skb
) {
984 if (more(buf_seqno(skb
), to
))
987 _skb
= __pskb_copy(skb
, MIN_H_SIZE
, GFP_ATOMIC
);
991 msg_set_ack(hdr
, ack
);
992 msg_set_bcast_ack(hdr
, bc_ack
);
993 _skb
->priority
= TC_PRIO_CONTROL
;
994 __skb_queue_tail(xmitq
, _skb
);
995 l
->stats
.retransmitted
++;
1000 /* tipc_data_input - deliver data and name distr msgs to upper layer
1002 * Consumes buffer if message is of right type
1003 * Node lock must be held
1005 static bool tipc_data_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1006 struct sk_buff_head
*inputq
)
1008 switch (msg_user(buf_msg(skb
))) {
1009 case TIPC_LOW_IMPORTANCE
:
1010 case TIPC_MEDIUM_IMPORTANCE
:
1011 case TIPC_HIGH_IMPORTANCE
:
1012 case TIPC_CRITICAL_IMPORTANCE
:
1014 skb_queue_tail(inputq
, skb
);
1016 case NAME_DISTRIBUTOR
:
1017 l
->bc_rcvlink
->state
= LINK_ESTABLISHED
;
1018 skb_queue_tail(l
->namedq
, skb
);
1021 case TUNNEL_PROTOCOL
:
1022 case MSG_FRAGMENTER
:
1023 case BCAST_PROTOCOL
:
1026 pr_warn("Dropping received illegal msg type\n");
1032 /* tipc_link_input - process packet that has passed link protocol check
1036 static int tipc_link_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1037 struct sk_buff_head
*inputq
)
1039 struct tipc_msg
*hdr
= buf_msg(skb
);
1040 struct sk_buff
**reasm_skb
= &l
->reasm_buf
;
1041 struct sk_buff
*iskb
;
1042 struct sk_buff_head tmpq
;
1043 int usr
= msg_user(hdr
);
1048 if (unlikely(usr
== TUNNEL_PROTOCOL
)) {
1049 if (msg_type(hdr
) == SYNCH_MSG
) {
1050 __skb_queue_purge(&l
->deferdq
);
1053 if (!tipc_msg_extract(skb
, &iskb
, &ipos
))
1058 if (less(msg_seqno(hdr
), l
->drop_point
))
1060 if (tipc_data_input(l
, skb
, inputq
))
1062 usr
= msg_user(hdr
);
1063 reasm_skb
= &l
->failover_reasm_skb
;
1066 if (usr
== MSG_BUNDLER
) {
1067 skb_queue_head_init(&tmpq
);
1068 l
->stats
.recv_bundles
++;
1069 l
->stats
.recv_bundled
+= msg_msgcnt(hdr
);
1070 while (tipc_msg_extract(skb
, &iskb
, &pos
))
1071 tipc_data_input(l
, iskb
, &tmpq
);
1072 tipc_skb_queue_splice_tail(&tmpq
, inputq
);
1074 } else if (usr
== MSG_FRAGMENTER
) {
1075 l
->stats
.recv_fragments
++;
1076 if (tipc_buf_append(reasm_skb
, &skb
)) {
1077 l
->stats
.recv_fragmented
++;
1078 tipc_data_input(l
, skb
, inputq
);
1079 } else if (!*reasm_skb
&& !link_is_bc_rcvlink(l
)) {
1080 pr_warn_ratelimited("Unable to build fragment list\n");
1081 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1084 } else if (usr
== BCAST_PROTOCOL
) {
1085 tipc_bcast_lock(l
->net
);
1086 tipc_link_bc_init_rcv(l
->bc_rcvlink
, hdr
);
1087 tipc_bcast_unlock(l
->net
);
1094 static bool tipc_link_release_pkts(struct tipc_link
*l
, u16 acked
)
1096 bool released
= false;
1097 struct sk_buff
*skb
, *tmp
;
1099 skb_queue_walk_safe(&l
->transmq
, skb
, tmp
) {
1100 if (more(buf_seqno(skb
), acked
))
1102 __skb_unlink(skb
, &l
->transmq
);
1109 /* tipc_link_build_state_msg: prepare link state message for transmission
1111 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1112 * risk of ack storms towards the sender
1114 int tipc_link_build_state_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1119 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1120 if (link_is_bc_rcvlink(l
)) {
1121 if (((l
->rcv_nxt
^ tipc_own_addr(l
->net
)) & 0xf) != 0xf)
1124 return TIPC_LINK_SND_BC_ACK
;
1129 l
->stats
.sent_acks
++;
1130 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, xmitq
);
1134 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1136 void tipc_link_build_reset_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1138 int mtyp
= RESET_MSG
;
1139 struct sk_buff
*skb
;
1141 if (l
->state
== LINK_ESTABLISHING
)
1142 mtyp
= ACTIVATE_MSG
;
1144 tipc_link_build_proto_msg(l
, mtyp
, 0, 0, 0, 0, xmitq
);
1146 /* Inform peer that this endpoint is going down if applicable */
1147 skb
= skb_peek_tail(xmitq
);
1148 if (skb
&& (l
->state
== LINK_RESET
))
1149 msg_set_peer_stopping(buf_msg(skb
), 1);
1152 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1154 static void tipc_link_build_nack_msg(struct tipc_link
*l
,
1155 struct sk_buff_head
*xmitq
)
1157 u32 def_cnt
= ++l
->stats
.deferred_recv
;
1159 if (link_is_bc_rcvlink(l
))
1162 if ((skb_queue_len(&l
->deferdq
) == 1) || !(def_cnt
% TIPC_NACK_INTV
))
1163 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, xmitq
);
1166 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1167 * @l: the link that should handle the message
1169 * @xmitq: queue to place packets to be sent after this call
1171 int tipc_link_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1172 struct sk_buff_head
*xmitq
)
1174 struct sk_buff_head
*defq
= &l
->deferdq
;
1175 struct tipc_msg
*hdr
;
1176 u16 seqno
, rcv_nxt
, win_lim
;
1181 seqno
= msg_seqno(hdr
);
1182 rcv_nxt
= l
->rcv_nxt
;
1183 win_lim
= rcv_nxt
+ TIPC_MAX_LINK_WIN
;
1185 /* Verify and update link state */
1186 if (unlikely(msg_user(hdr
) == LINK_PROTOCOL
))
1187 return tipc_link_proto_rcv(l
, skb
, xmitq
);
1189 if (unlikely(!link_is_up(l
))) {
1190 if (l
->state
== LINK_ESTABLISHING
)
1191 rc
= TIPC_LINK_UP_EVT
;
1195 /* Don't send probe at next timeout expiration */
1196 l
->silent_intv_cnt
= 0;
1198 /* Drop if outside receive window */
1199 if (unlikely(less(seqno
, rcv_nxt
) || more(seqno
, win_lim
))) {
1200 l
->stats
.duplicates
++;
1204 /* Forward queues and wake up waiting users */
1205 if (likely(tipc_link_release_pkts(l
, msg_ack(hdr
)))) {
1206 tipc_link_advance_backlog(l
, xmitq
);
1207 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
1208 link_prepare_wakeup(l
);
1211 /* Defer delivery if sequence gap */
1212 if (unlikely(seqno
!= rcv_nxt
)) {
1213 __tipc_skb_queue_sorted(defq
, seqno
, skb
);
1214 tipc_link_build_nack_msg(l
, xmitq
);
1218 /* Deliver packet */
1220 l
->stats
.recv_info
++;
1221 if (!tipc_data_input(l
, skb
, l
->inputq
))
1222 rc
|= tipc_link_input(l
, skb
, l
->inputq
);
1223 if (unlikely(++l
->rcv_unacked
>= TIPC_MIN_LINK_WIN
))
1224 rc
|= tipc_link_build_state_msg(l
, xmitq
);
1225 if (unlikely(rc
& ~TIPC_LINK_SND_BC_ACK
))
1227 } while ((skb
= __skb_dequeue(defq
)));
1235 static void tipc_link_build_proto_msg(struct tipc_link
*l
, int mtyp
, bool probe
,
1236 u16 rcvgap
, int tolerance
, int priority
,
1237 struct sk_buff_head
*xmitq
)
1239 struct sk_buff
*skb
;
1240 struct tipc_msg
*hdr
;
1241 struct sk_buff_head
*dfq
= &l
->deferdq
;
1242 bool node_up
= link_is_up(l
->bc_rcvlink
);
1244 /* Don't send protocol message during reset or link failover */
1245 if (tipc_link_is_blocked(l
))
1248 if (!tipc_link_is_up(l
) && (mtyp
== STATE_MSG
))
1251 if (!skb_queue_empty(dfq
))
1252 rcvgap
= buf_seqno(skb_peek(dfq
)) - l
->rcv_nxt
;
1254 skb
= tipc_msg_create(LINK_PROTOCOL
, mtyp
, INT_H_SIZE
,
1255 TIPC_MAX_IF_NAME
, l
->addr
,
1256 tipc_own_addr(l
->net
), 0, 0, 0);
1261 msg_set_session(hdr
, l
->session
);
1262 msg_set_bearer_id(hdr
, l
->bearer_id
);
1263 msg_set_net_plane(hdr
, l
->net_plane
);
1264 msg_set_next_sent(hdr
, l
->snd_nxt
);
1265 msg_set_ack(hdr
, l
->rcv_nxt
- 1);
1266 msg_set_bcast_ack(hdr
, l
->bc_rcvlink
->rcv_nxt
- 1);
1267 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
1268 msg_set_link_tolerance(hdr
, tolerance
);
1269 msg_set_linkprio(hdr
, priority
);
1270 msg_set_redundant_link(hdr
, node_up
);
1271 msg_set_seq_gap(hdr
, 0);
1272 msg_set_seqno(hdr
, l
->snd_nxt
+ U16_MAX
/ 2);
1274 if (mtyp
== STATE_MSG
) {
1275 msg_set_seq_gap(hdr
, rcvgap
);
1276 msg_set_size(hdr
, INT_H_SIZE
);
1277 msg_set_probe(hdr
, probe
);
1278 l
->stats
.sent_states
++;
1281 /* RESET_MSG or ACTIVATE_MSG */
1282 msg_set_max_pkt(hdr
, l
->advertised_mtu
);
1283 strcpy(msg_data(hdr
), l
->if_name
);
1286 l
->stats
.sent_probes
++;
1288 l
->stats
.sent_nacks
++;
1289 skb
->priority
= TC_PRIO_CONTROL
;
1290 __skb_queue_tail(xmitq
, skb
);
1293 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1294 * with contents of the link's transmit and backlog queues.
1296 void tipc_link_tnl_prepare(struct tipc_link
*l
, struct tipc_link
*tnl
,
1297 int mtyp
, struct sk_buff_head
*xmitq
)
1299 struct sk_buff
*skb
, *tnlskb
;
1300 struct tipc_msg
*hdr
, tnlhdr
;
1301 struct sk_buff_head
*queue
= &l
->transmq
;
1302 struct sk_buff_head tmpxq
, tnlq
;
1303 u16 pktlen
, pktcnt
, seqno
= l
->snd_nxt
;
1308 skb_queue_head_init(&tnlq
);
1309 skb_queue_head_init(&tmpxq
);
1311 /* At least one packet required for safe algorithm => add dummy */
1312 skb
= tipc_msg_create(TIPC_LOW_IMPORTANCE
, TIPC_DIRECT_MSG
,
1313 BASIC_H_SIZE
, 0, l
->addr
, tipc_own_addr(l
->net
),
1314 0, 0, TIPC_ERR_NO_PORT
);
1316 pr_warn("%sunable to create tunnel packet\n", link_co_err
);
1319 skb_queue_tail(&tnlq
, skb
);
1320 tipc_link_xmit(l
, &tnlq
, &tmpxq
);
1321 __skb_queue_purge(&tmpxq
);
1323 /* Initialize reusable tunnel packet header */
1324 tipc_msg_init(tipc_own_addr(l
->net
), &tnlhdr
, TUNNEL_PROTOCOL
,
1325 mtyp
, INT_H_SIZE
, l
->addr
);
1326 pktcnt
= skb_queue_len(&l
->transmq
) + skb_queue_len(&l
->backlogq
);
1327 msg_set_msgcnt(&tnlhdr
, pktcnt
);
1328 msg_set_bearer_id(&tnlhdr
, l
->peer_bearer_id
);
1330 /* Wrap each packet into a tunnel packet */
1331 skb_queue_walk(queue
, skb
) {
1333 if (queue
== &l
->backlogq
)
1334 msg_set_seqno(hdr
, seqno
++);
1335 pktlen
= msg_size(hdr
);
1336 msg_set_size(&tnlhdr
, pktlen
+ INT_H_SIZE
);
1337 tnlskb
= tipc_buf_acquire(pktlen
+ INT_H_SIZE
);
1339 pr_warn("%sunable to send packet\n", link_co_err
);
1342 skb_copy_to_linear_data(tnlskb
, &tnlhdr
, INT_H_SIZE
);
1343 skb_copy_to_linear_data_offset(tnlskb
, INT_H_SIZE
, hdr
, pktlen
);
1344 __skb_queue_tail(&tnlq
, tnlskb
);
1346 if (queue
!= &l
->backlogq
) {
1347 queue
= &l
->backlogq
;
1351 tipc_link_xmit(tnl
, &tnlq
, xmitq
);
1353 if (mtyp
== FAILOVER_MSG
) {
1354 tnl
->drop_point
= l
->rcv_nxt
;
1355 tnl
->failover_reasm_skb
= l
->reasm_buf
;
1356 l
->reasm_buf
= NULL
;
1360 /* tipc_link_proto_rcv(): receive link level protocol message :
1361 * Note that network plane id propagates through the network, and may
1362 * change at any time. The node with lowest numerical id determines
1365 static int tipc_link_proto_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1366 struct sk_buff_head
*xmitq
)
1368 struct tipc_msg
*hdr
= buf_msg(skb
);
1370 u16 ack
= msg_ack(hdr
);
1371 u16 gap
= msg_seq_gap(hdr
);
1372 u16 peers_snd_nxt
= msg_next_sent(hdr
);
1373 u16 peers_tol
= msg_link_tolerance(hdr
);
1374 u16 peers_prio
= msg_linkprio(hdr
);
1375 u16 rcv_nxt
= l
->rcv_nxt
;
1376 int mtyp
= msg_type(hdr
);
1380 if (tipc_link_is_blocked(l
) || !xmitq
)
1383 if (tipc_own_addr(l
->net
) > msg_prevnode(hdr
))
1384 l
->net_plane
= msg_net_plane(hdr
);
1389 /* Ignore duplicate RESET with old session number */
1390 if ((less_eq(msg_session(hdr
), l
->peer_session
)) &&
1391 (l
->peer_session
!= ANY_SESSION
))
1399 /* Complete own link name with peer's interface name */
1400 if_name
= strrchr(l
->name
, ':') + 1;
1401 if (sizeof(l
->name
) - (if_name
- l
->name
) <= TIPC_MAX_IF_NAME
)
1403 if (msg_data_sz(hdr
) < TIPC_MAX_IF_NAME
)
1405 strncpy(if_name
, msg_data(hdr
), TIPC_MAX_IF_NAME
);
1407 /* Update own tolerance if peer indicates a non-zero value */
1408 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
))
1409 l
->tolerance
= peers_tol
;
1411 /* Update own priority if peer's priority is higher */
1412 if (in_range(peers_prio
, l
->priority
+ 1, TIPC_MAX_LINK_PRI
))
1413 l
->priority
= peers_prio
;
1415 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1416 if (msg_peer_stopping(hdr
))
1417 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1418 else if ((mtyp
== RESET_MSG
) || !link_is_up(l
))
1419 rc
= tipc_link_fsm_evt(l
, LINK_PEER_RESET_EVT
);
1421 /* ACTIVATE_MSG takes up link if it was already locally reset */
1422 if ((mtyp
== ACTIVATE_MSG
) && (l
->state
== LINK_ESTABLISHING
))
1423 rc
= TIPC_LINK_UP_EVT
;
1425 l
->peer_session
= msg_session(hdr
);
1426 l
->peer_bearer_id
= msg_bearer_id(hdr
);
1427 if (l
->mtu
> msg_max_pkt(hdr
))
1428 l
->mtu
= msg_max_pkt(hdr
);
1433 /* Update own tolerance if peer indicates a non-zero value */
1434 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
))
1435 l
->tolerance
= peers_tol
;
1437 if (peers_prio
&& in_range(peers_prio
, TIPC_MIN_LINK_PRI
,
1438 TIPC_MAX_LINK_PRI
)) {
1439 l
->priority
= peers_prio
;
1440 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1443 l
->silent_intv_cnt
= 0;
1444 l
->stats
.recv_states
++;
1446 l
->stats
.recv_probes
++;
1448 if (!link_is_up(l
)) {
1449 if (l
->state
== LINK_ESTABLISHING
)
1450 rc
= TIPC_LINK_UP_EVT
;
1454 /* Send NACK if peer has sent pkts we haven't received yet */
1455 if (more(peers_snd_nxt
, rcv_nxt
) && !tipc_link_is_synching(l
))
1456 rcvgap
= peers_snd_nxt
- l
->rcv_nxt
;
1457 if (rcvgap
|| (msg_probe(hdr
)))
1458 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, rcvgap
,
1460 tipc_link_release_pkts(l
, ack
);
1462 /* If NACK, retransmit will now start at right position */
1464 rc
= tipc_link_retrans(l
, ack
+ 1, ack
+ gap
, xmitq
);
1465 l
->stats
.recv_nacks
++;
1468 tipc_link_advance_backlog(l
, xmitq
);
1469 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
1470 link_prepare_wakeup(l
);
1477 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1479 static bool tipc_link_build_bc_proto_msg(struct tipc_link
*l
, bool bcast
,
1481 struct sk_buff_head
*xmitq
)
1483 struct sk_buff
*skb
;
1484 struct tipc_msg
*hdr
;
1485 struct sk_buff
*dfrd_skb
= skb_peek(&l
->deferdq
);
1486 u16 ack
= l
->rcv_nxt
- 1;
1487 u16 gap_to
= peers_snd_nxt
- 1;
1489 skb
= tipc_msg_create(BCAST_PROTOCOL
, STATE_MSG
, INT_H_SIZE
,
1490 0, l
->addr
, tipc_own_addr(l
->net
), 0, 0, 0);
1494 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
1495 msg_set_bcast_ack(hdr
, ack
);
1496 msg_set_bcgap_after(hdr
, ack
);
1498 gap_to
= buf_seqno(dfrd_skb
) - 1;
1499 msg_set_bcgap_to(hdr
, gap_to
);
1500 msg_set_non_seq(hdr
, bcast
);
1501 __skb_queue_tail(xmitq
, skb
);
1505 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1507 * Give a newly added peer node the sequence number where it should
1508 * start receiving and acking broadcast packets.
1510 static void tipc_link_build_bc_init_msg(struct tipc_link
*l
,
1511 struct sk_buff_head
*xmitq
)
1513 struct sk_buff_head list
;
1515 __skb_queue_head_init(&list
);
1516 if (!tipc_link_build_bc_proto_msg(l
->bc_rcvlink
, false, 0, &list
))
1518 tipc_link_xmit(l
, &list
, xmitq
);
1521 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1523 void tipc_link_bc_init_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
)
1525 int mtyp
= msg_type(hdr
);
1526 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
1531 if (msg_user(hdr
) == BCAST_PROTOCOL
) {
1532 l
->rcv_nxt
= peers_snd_nxt
;
1533 l
->state
= LINK_ESTABLISHED
;
1537 if (l
->peer_caps
& TIPC_BCAST_SYNCH
)
1540 if (msg_peer_node_is_up(hdr
))
1543 /* Compatibility: accept older, less safe initial synch data */
1544 if ((mtyp
== RESET_MSG
) || (mtyp
== ACTIVATE_MSG
))
1545 l
->rcv_nxt
= peers_snd_nxt
;
1548 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1550 void tipc_link_bc_sync_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
,
1551 struct sk_buff_head
*xmitq
)
1553 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
1558 if (!msg_peer_node_is_up(hdr
))
1561 l
->bc_peer_is_up
= true;
1563 /* Ignore if peers_snd_nxt goes beyond receive window */
1564 if (more(peers_snd_nxt
, l
->rcv_nxt
+ l
->window
))
1567 if (!more(peers_snd_nxt
, l
->rcv_nxt
)) {
1568 l
->nack_state
= BC_NACK_SND_CONDITIONAL
;
1572 /* Don't NACK if one was recently sent or peeked */
1573 if (l
->nack_state
== BC_NACK_SND_SUPPRESS
) {
1574 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
1578 /* Conditionally delay NACK sending until next synch rcv */
1579 if (l
->nack_state
== BC_NACK_SND_CONDITIONAL
) {
1580 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
1581 if ((peers_snd_nxt
- l
->rcv_nxt
) < TIPC_MIN_LINK_WIN
)
1585 /* Send NACK now but suppress next one */
1586 tipc_link_build_bc_proto_msg(l
, true, peers_snd_nxt
, xmitq
);
1587 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
1590 void tipc_link_bc_ack_rcv(struct tipc_link
*l
, u16 acked
,
1591 struct sk_buff_head
*xmitq
)
1593 struct sk_buff
*skb
, *tmp
;
1594 struct tipc_link
*snd_l
= l
->bc_sndlink
;
1596 if (!link_is_up(l
) || !l
->bc_peer_is_up
)
1599 if (!more(acked
, l
->acked
))
1602 /* Skip over packets peer has already acked */
1603 skb_queue_walk(&snd_l
->transmq
, skb
) {
1604 if (more(buf_seqno(skb
), l
->acked
))
1608 /* Update/release the packets peer is acking now */
1609 skb_queue_walk_from_safe(&snd_l
->transmq
, skb
, tmp
) {
1610 if (more(buf_seqno(skb
), acked
))
1612 if (!--TIPC_SKB_CB(skb
)->ackers
) {
1613 __skb_unlink(skb
, &snd_l
->transmq
);
1618 tipc_link_advance_backlog(snd_l
, xmitq
);
1619 if (unlikely(!skb_queue_empty(&snd_l
->wakeupq
)))
1620 link_prepare_wakeup(snd_l
);
1623 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1625 int tipc_link_bc_nack_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1626 struct sk_buff_head
*xmitq
)
1628 struct tipc_msg
*hdr
= buf_msg(skb
);
1629 u32 dnode
= msg_destnode(hdr
);
1630 int mtyp
= msg_type(hdr
);
1631 u16 acked
= msg_bcast_ack(hdr
);
1632 u16 from
= acked
+ 1;
1633 u16 to
= msg_bcgap_to(hdr
);
1634 u16 peers_snd_nxt
= to
+ 1;
1639 if (!tipc_link_is_up(l
) || !l
->bc_peer_is_up
)
1642 if (mtyp
!= STATE_MSG
)
1645 if (dnode
== tipc_own_addr(l
->net
)) {
1646 tipc_link_bc_ack_rcv(l
, acked
, xmitq
);
1647 rc
= tipc_link_retrans(l
->bc_sndlink
, from
, to
, xmitq
);
1648 l
->stats
.recv_nacks
++;
1652 /* Msg for other node => suppress own NACK at next sync if applicable */
1653 if (more(peers_snd_nxt
, l
->rcv_nxt
) && !less(l
->rcv_nxt
, from
))
1654 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
1659 void tipc_link_set_queue_limits(struct tipc_link
*l
, u32 win
)
1661 int max_bulk
= TIPC_MAX_PUBLICATIONS
/ (l
->mtu
/ ITEM_SIZE
);
1664 l
->backlog
[TIPC_LOW_IMPORTANCE
].limit
= win
/ 2;
1665 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].limit
= win
;
1666 l
->backlog
[TIPC_HIGH_IMPORTANCE
].limit
= win
/ 2 * 3;
1667 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].limit
= win
* 2;
1668 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].limit
= max_bulk
;
1672 * link_reset_stats - reset link statistics
1673 * @l: pointer to link
1675 void tipc_link_reset_stats(struct tipc_link
*l
)
1677 memset(&l
->stats
, 0, sizeof(l
->stats
));
1678 if (!link_is_bc_sndlink(l
)) {
1679 l
->stats
.sent_info
= l
->snd_nxt
;
1680 l
->stats
.recv_info
= l
->rcv_nxt
;
1684 static void link_print(struct tipc_link
*l
, const char *str
)
1686 struct sk_buff
*hskb
= skb_peek(&l
->transmq
);
1687 u16 head
= hskb
? msg_seqno(buf_msg(hskb
)) : l
->snd_nxt
- 1;
1688 u16 tail
= l
->snd_nxt
- 1;
1690 pr_info("%s Link <%s> state %x\n", str
, l
->name
, l
->state
);
1691 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1692 skb_queue_len(&l
->transmq
), head
, tail
,
1693 skb_queue_len(&l
->backlogq
), l
->snd_nxt
, l
->rcv_nxt
);
1696 /* Parse and validate nested (link) properties valid for media, bearer and link
1698 int tipc_nl_parse_link_prop(struct nlattr
*prop
, struct nlattr
*props
[])
1702 err
= nla_parse_nested(props
, TIPC_NLA_PROP_MAX
, prop
,
1703 tipc_nl_prop_policy
);
1707 if (props
[TIPC_NLA_PROP_PRIO
]) {
1710 prio
= nla_get_u32(props
[TIPC_NLA_PROP_PRIO
]);
1711 if (prio
> TIPC_MAX_LINK_PRI
)
1715 if (props
[TIPC_NLA_PROP_TOL
]) {
1718 tol
= nla_get_u32(props
[TIPC_NLA_PROP_TOL
]);
1719 if ((tol
< TIPC_MIN_LINK_TOL
) || (tol
> TIPC_MAX_LINK_TOL
))
1723 if (props
[TIPC_NLA_PROP_WIN
]) {
1726 win
= nla_get_u32(props
[TIPC_NLA_PROP_WIN
]);
1727 if ((win
< TIPC_MIN_LINK_WIN
) || (win
> TIPC_MAX_LINK_WIN
))
1734 static int __tipc_nl_add_stats(struct sk_buff
*skb
, struct tipc_stats
*s
)
1737 struct nlattr
*stats
;
1744 struct nla_map map
[] = {
1745 {TIPC_NLA_STATS_RX_INFO
, s
->recv_info
},
1746 {TIPC_NLA_STATS_RX_FRAGMENTS
, s
->recv_fragments
},
1747 {TIPC_NLA_STATS_RX_FRAGMENTED
, s
->recv_fragmented
},
1748 {TIPC_NLA_STATS_RX_BUNDLES
, s
->recv_bundles
},
1749 {TIPC_NLA_STATS_RX_BUNDLED
, s
->recv_bundled
},
1750 {TIPC_NLA_STATS_TX_INFO
, s
->sent_info
},
1751 {TIPC_NLA_STATS_TX_FRAGMENTS
, s
->sent_fragments
},
1752 {TIPC_NLA_STATS_TX_FRAGMENTED
, s
->sent_fragmented
},
1753 {TIPC_NLA_STATS_TX_BUNDLES
, s
->sent_bundles
},
1754 {TIPC_NLA_STATS_TX_BUNDLED
, s
->sent_bundled
},
1755 {TIPC_NLA_STATS_MSG_PROF_TOT
, (s
->msg_length_counts
) ?
1756 s
->msg_length_counts
: 1},
1757 {TIPC_NLA_STATS_MSG_LEN_CNT
, s
->msg_length_counts
},
1758 {TIPC_NLA_STATS_MSG_LEN_TOT
, s
->msg_lengths_total
},
1759 {TIPC_NLA_STATS_MSG_LEN_P0
, s
->msg_length_profile
[0]},
1760 {TIPC_NLA_STATS_MSG_LEN_P1
, s
->msg_length_profile
[1]},
1761 {TIPC_NLA_STATS_MSG_LEN_P2
, s
->msg_length_profile
[2]},
1762 {TIPC_NLA_STATS_MSG_LEN_P3
, s
->msg_length_profile
[3]},
1763 {TIPC_NLA_STATS_MSG_LEN_P4
, s
->msg_length_profile
[4]},
1764 {TIPC_NLA_STATS_MSG_LEN_P5
, s
->msg_length_profile
[5]},
1765 {TIPC_NLA_STATS_MSG_LEN_P6
, s
->msg_length_profile
[6]},
1766 {TIPC_NLA_STATS_RX_STATES
, s
->recv_states
},
1767 {TIPC_NLA_STATS_RX_PROBES
, s
->recv_probes
},
1768 {TIPC_NLA_STATS_RX_NACKS
, s
->recv_nacks
},
1769 {TIPC_NLA_STATS_RX_DEFERRED
, s
->deferred_recv
},
1770 {TIPC_NLA_STATS_TX_STATES
, s
->sent_states
},
1771 {TIPC_NLA_STATS_TX_PROBES
, s
->sent_probes
},
1772 {TIPC_NLA_STATS_TX_NACKS
, s
->sent_nacks
},
1773 {TIPC_NLA_STATS_TX_ACKS
, s
->sent_acks
},
1774 {TIPC_NLA_STATS_RETRANSMITTED
, s
->retransmitted
},
1775 {TIPC_NLA_STATS_DUPLICATES
, s
->duplicates
},
1776 {TIPC_NLA_STATS_LINK_CONGS
, s
->link_congs
},
1777 {TIPC_NLA_STATS_MAX_QUEUE
, s
->max_queue_sz
},
1778 {TIPC_NLA_STATS_AVG_QUEUE
, s
->queue_sz_counts
?
1779 (s
->accu_queue_sz
/ s
->queue_sz_counts
) : 0}
1782 stats
= nla_nest_start(skb
, TIPC_NLA_LINK_STATS
);
1786 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
1787 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
1790 nla_nest_end(skb
, stats
);
1794 nla_nest_cancel(skb
, stats
);
1799 /* Caller should hold appropriate locks to protect the link */
1800 int __tipc_nl_add_link(struct net
*net
, struct tipc_nl_msg
*msg
,
1801 struct tipc_link
*link
, int nlflags
)
1805 struct nlattr
*attrs
;
1806 struct nlattr
*prop
;
1807 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
1809 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
1810 nlflags
, TIPC_NL_LINK_GET
);
1814 attrs
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK
);
1818 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, link
->name
))
1820 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_DEST
,
1821 tipc_cluster_mask(tn
->own_addr
)))
1823 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_MTU
, link
->mtu
))
1825 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, link
->rcv_nxt
))
1827 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, link
->snd_nxt
))
1830 if (tipc_link_is_up(link
))
1831 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
1834 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_ACTIVE
))
1837 prop
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK_PROP
);
1840 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
1842 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_TOL
, link
->tolerance
))
1844 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
,
1847 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
1849 nla_nest_end(msg
->skb
, prop
);
1851 err
= __tipc_nl_add_stats(msg
->skb
, &link
->stats
);
1855 nla_nest_end(msg
->skb
, attrs
);
1856 genlmsg_end(msg
->skb
, hdr
);
1861 nla_nest_cancel(msg
->skb
, prop
);
1863 nla_nest_cancel(msg
->skb
, attrs
);
1865 genlmsg_cancel(msg
->skb
, hdr
);
1870 static int __tipc_nl_add_bc_link_stat(struct sk_buff
*skb
,
1871 struct tipc_stats
*stats
)
1874 struct nlattr
*nest
;
1881 struct nla_map map
[] = {
1882 {TIPC_NLA_STATS_RX_INFO
, stats
->recv_info
},
1883 {TIPC_NLA_STATS_RX_FRAGMENTS
, stats
->recv_fragments
},
1884 {TIPC_NLA_STATS_RX_FRAGMENTED
, stats
->recv_fragmented
},
1885 {TIPC_NLA_STATS_RX_BUNDLES
, stats
->recv_bundles
},
1886 {TIPC_NLA_STATS_RX_BUNDLED
, stats
->recv_bundled
},
1887 {TIPC_NLA_STATS_TX_INFO
, stats
->sent_info
},
1888 {TIPC_NLA_STATS_TX_FRAGMENTS
, stats
->sent_fragments
},
1889 {TIPC_NLA_STATS_TX_FRAGMENTED
, stats
->sent_fragmented
},
1890 {TIPC_NLA_STATS_TX_BUNDLES
, stats
->sent_bundles
},
1891 {TIPC_NLA_STATS_TX_BUNDLED
, stats
->sent_bundled
},
1892 {TIPC_NLA_STATS_RX_NACKS
, stats
->recv_nacks
},
1893 {TIPC_NLA_STATS_RX_DEFERRED
, stats
->deferred_recv
},
1894 {TIPC_NLA_STATS_TX_NACKS
, stats
->sent_nacks
},
1895 {TIPC_NLA_STATS_TX_ACKS
, stats
->sent_acks
},
1896 {TIPC_NLA_STATS_RETRANSMITTED
, stats
->retransmitted
},
1897 {TIPC_NLA_STATS_DUPLICATES
, stats
->duplicates
},
1898 {TIPC_NLA_STATS_LINK_CONGS
, stats
->link_congs
},
1899 {TIPC_NLA_STATS_MAX_QUEUE
, stats
->max_queue_sz
},
1900 {TIPC_NLA_STATS_AVG_QUEUE
, stats
->queue_sz_counts
?
1901 (stats
->accu_queue_sz
/ stats
->queue_sz_counts
) : 0}
1904 nest
= nla_nest_start(skb
, TIPC_NLA_LINK_STATS
);
1908 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
1909 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
1912 nla_nest_end(skb
, nest
);
1916 nla_nest_cancel(skb
, nest
);
1921 int tipc_nl_add_bc_link(struct net
*net
, struct tipc_nl_msg
*msg
)
1925 struct nlattr
*attrs
;
1926 struct nlattr
*prop
;
1927 struct tipc_net
*tn
= net_generic(net
, tipc_net_id
);
1928 struct tipc_link
*bcl
= tn
->bcl
;
1933 tipc_bcast_lock(net
);
1935 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
1936 NLM_F_MULTI
, TIPC_NL_LINK_GET
);
1938 tipc_bcast_unlock(net
);
1942 attrs
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK
);
1946 /* The broadcast link is always up */
1947 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
1950 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_BROADCAST
))
1952 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, bcl
->name
))
1954 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, bcl
->rcv_nxt
))
1956 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, bcl
->snd_nxt
))
1959 prop
= nla_nest_start(msg
->skb
, TIPC_NLA_LINK_PROP
);
1962 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
, bcl
->window
))
1964 nla_nest_end(msg
->skb
, prop
);
1966 err
= __tipc_nl_add_bc_link_stat(msg
->skb
, &bcl
->stats
);
1970 tipc_bcast_unlock(net
);
1971 nla_nest_end(msg
->skb
, attrs
);
1972 genlmsg_end(msg
->skb
, hdr
);
1977 nla_nest_cancel(msg
->skb
, prop
);
1979 nla_nest_cancel(msg
->skb
, attrs
);
1981 tipc_bcast_unlock(net
);
1982 genlmsg_cancel(msg
->skb
, hdr
);
1987 void tipc_link_set_tolerance(struct tipc_link
*l
, u32 tol
,
1988 struct sk_buff_head
*xmitq
)
1991 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, tol
, 0, xmitq
);
1994 void tipc_link_set_prio(struct tipc_link
*l
, u32 prio
,
1995 struct sk_buff_head
*xmitq
)
1998 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, prio
, xmitq
);
2001 void tipc_link_set_abort_limit(struct tipc_link
*l
, u32 limit
)
2003 l
->abort_limit
= limit
;