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"
49 #include <linux/pkt_sched.h>
70 u32 link_congs
; /* # port sends blocked by congestion */
73 u32 max_queue_sz
; /* send queue size high water mark */
74 u32 accu_queue_sz
; /* used for send queue size profiling */
75 u32 queue_sz_counts
; /* used for send queue size profiling */
76 u32 msg_length_counts
; /* used for message length profiling */
77 u32 msg_lengths_total
; /* used for message length profiling */
78 u32 msg_length_profile
[7]; /* used for msg. length profiling */
82 * struct tipc_link - TIPC link data structure
83 * @addr: network address of link's peer node
84 * @name: link name character string
85 * @media_addr: media address to use when sending messages over link
87 * @net: pointer to namespace struct
88 * @refcnt: reference counter for permanent references (owner node & timer)
89 * @peer_session: link session # being used by peer end of link
90 * @peer_bearer_id: bearer id used by link's peer endpoint
91 * @bearer_id: local bearer id used by link
92 * @tolerance: minimum link continuity loss needed to reset link [in ms]
93 * @abort_limit: # of unacknowledged continuity probes needed to reset link
94 * @state: current state of link FSM
95 * @peer_caps: bitmap describing capabilities of peer node
96 * @silent_intv_cnt: # of timer intervals without any reception from peer
97 * @proto_msg: template for control messages generated by link
98 * @pmsg: convenience pointer to "proto_msg" field
99 * @priority: current link priority
100 * @net_plane: current link network plane ('A' through 'H')
101 * @mon_state: cookie with information needed by link monitor
102 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
103 * @exp_msg_count: # of tunnelled messages expected during link changeover
104 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
105 * @mtu: current maximum packet size for this link
106 * @advertised_mtu: advertised own mtu when link is being established
107 * @transmitq: queue for sent, non-acked messages
108 * @backlogq: queue for messages waiting to be sent
109 * @snt_nxt: next sequence number to use for outbound messages
110 * @ackers: # of peers that needs to ack each packet before it can be released
111 * @acked: # last packet acked by a certain peer. Used for broadcast.
112 * @rcv_nxt: next sequence number to expect for inbound messages
113 * @deferred_queue: deferred queue saved OOS b'cast message received from node
114 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115 * @inputq: buffer queue for messages to be delivered upwards
116 * @namedq: buffer queue for name table messages to be delivered upwards
117 * @next_out: ptr to first unsent outbound message in queue
118 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120 * @reasm_buf: head of partially reassembled inbound message fragments
121 * @bc_rcvr: marks that this is a broadcast receiver link
122 * @stats: collects statistics regarding link activity
123 * @session: session to be used by link
124 * @snd_nxt_state: next send seq number
125 * @rcv_nxt_state: next rcv seq number
126 * @in_session: have received ACTIVATE_MSG from peer
127 * @active: link is active
128 * @if_name: associated interface name
129 * @rst_cnt: link reset counter
130 * @drop_point: seq number for failover handling (FIXME)
131 * @failover_reasm_skb: saved failover msg ptr (FIXME)
132 * @failover_deferdq: deferred message queue for failover processing (FIXME)
133 * @transmq: the link's transmit queue
134 * @backlog: link's backlog by priority (importance)
135 * @snd_nxt: next sequence number to be used
136 * @rcv_unacked: # messages read by user, but not yet acked back to peer
137 * @deferdq: deferred receive queue
138 * @window: sliding window size for congestion handling
139 * @min_win: minimal send window to be used by link
140 * @ssthresh: slow start threshold for congestion handling
141 * @max_win: maximal send window to be used by link
142 * @cong_acks: congestion acks for congestion avoidance (FIXME)
143 * @checkpoint: seq number for congestion window size handling
144 * @reasm_tnlmsg: fragmentation/reassembly area for tunnel protocol message
145 * @last_gap: last gap ack blocks for bcast (FIXME)
146 * @last_ga: ptr to gap ack blocks
147 * @bc_rcvlink: the peer specific link used for broadcast reception
148 * @bc_sndlink: the namespace global link used for broadcast sending
149 * @nack_state: bcast nack state
150 * @bc_peer_is_up: peer has acked the bcast init msg
154 char name
[TIPC_MAX_LINK_NAME
];
157 /* Management and link supervision data */
171 char if_name
[TIPC_MAX_IF_NAME
];
174 struct tipc_mon_state mon_state
;
179 struct sk_buff
*failover_reasm_skb
;
180 struct sk_buff_head failover_deferdq
;
182 /* Max packet negotiation */
187 struct sk_buff_head transmq
;
188 struct sk_buff_head backlogq
;
192 struct sk_buff
*target_bskb
;
199 struct sk_buff_head deferdq
;
200 struct sk_buff_head
*inputq
;
201 struct sk_buff_head
*namedq
;
203 /* Congestion handling */
204 struct sk_buff_head wakeupq
;
212 /* Fragmentation/reassembly */
213 struct sk_buff
*reasm_buf
;
214 struct sk_buff
*reasm_tnlmsg
;
220 struct tipc_gap_ack_blks
*last_ga
;
221 struct tipc_link
*bc_rcvlink
;
222 struct tipc_link
*bc_sndlink
;
227 struct tipc_stats stats
;
231 * Error message prefixes
233 static const char *link_co_err
= "Link tunneling error, ";
234 static const char *link_rst_msg
= "Resetting link ";
236 /* Send states for broadcast NACKs
239 BC_NACK_SND_CONDITIONAL
,
240 BC_NACK_SND_UNCONDITIONAL
,
241 BC_NACK_SND_SUPPRESS
,
244 #define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
245 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
250 LINK_ESTABLISHED
= 0xe,
251 LINK_ESTABLISHING
= 0xe << 4,
252 LINK_RESET
= 0x1 << 8,
253 LINK_RESETTING
= 0x2 << 12,
254 LINK_PEER_RESET
= 0xd << 16,
255 LINK_FAILINGOVER
= 0xf << 20,
256 LINK_SYNCHING
= 0xc << 24
259 /* Link FSM state checking routines
261 static int link_is_up(struct tipc_link
*l
)
263 return l
->state
& (LINK_ESTABLISHED
| LINK_SYNCHING
);
266 static int tipc_link_proto_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
267 struct sk_buff_head
*xmitq
);
268 static void tipc_link_build_proto_msg(struct tipc_link
*l
, int mtyp
, bool probe
,
269 bool probe_reply
, u16 rcvgap
,
270 int tolerance
, int priority
,
271 struct sk_buff_head
*xmitq
);
272 static void link_print(struct tipc_link
*l
, const char *str
);
273 static int tipc_link_build_nack_msg(struct tipc_link
*l
,
274 struct sk_buff_head
*xmitq
);
275 static void tipc_link_build_bc_init_msg(struct tipc_link
*l
,
276 struct sk_buff_head
*xmitq
);
277 static u8
__tipc_build_gap_ack_blks(struct tipc_gap_ack_blks
*ga
,
278 struct tipc_link
*l
, u8 start_index
);
279 static u16
tipc_build_gap_ack_blks(struct tipc_link
*l
, struct tipc_msg
*hdr
);
280 static int tipc_link_advance_transmq(struct tipc_link
*l
, struct tipc_link
*r
,
282 struct tipc_gap_ack_blks
*ga
,
283 struct sk_buff_head
*xmitq
,
284 bool *retransmitted
, int *rc
);
285 static void tipc_link_update_cwin(struct tipc_link
*l
, int released
,
288 * Simple non-static link routines (i.e. referenced outside this file)
290 bool tipc_link_is_up(struct tipc_link
*l
)
292 return link_is_up(l
);
295 bool tipc_link_peer_is_down(struct tipc_link
*l
)
297 return l
->state
== LINK_PEER_RESET
;
300 bool tipc_link_is_reset(struct tipc_link
*l
)
302 return l
->state
& (LINK_RESET
| LINK_FAILINGOVER
| LINK_ESTABLISHING
);
305 bool tipc_link_is_establishing(struct tipc_link
*l
)
307 return l
->state
== LINK_ESTABLISHING
;
310 bool tipc_link_is_synching(struct tipc_link
*l
)
312 return l
->state
== LINK_SYNCHING
;
315 bool tipc_link_is_failingover(struct tipc_link
*l
)
317 return l
->state
== LINK_FAILINGOVER
;
320 bool tipc_link_is_blocked(struct tipc_link
*l
)
322 return l
->state
& (LINK_RESETTING
| LINK_PEER_RESET
| LINK_FAILINGOVER
);
325 static bool link_is_bc_sndlink(struct tipc_link
*l
)
327 return !l
->bc_sndlink
;
330 static bool link_is_bc_rcvlink(struct tipc_link
*l
)
332 return ((l
->bc_rcvlink
== l
) && !link_is_bc_sndlink(l
));
335 void tipc_link_set_active(struct tipc_link
*l
, bool active
)
340 u32
tipc_link_id(struct tipc_link
*l
)
342 return l
->peer_bearer_id
<< 16 | l
->bearer_id
;
345 int tipc_link_min_win(struct tipc_link
*l
)
350 int tipc_link_max_win(struct tipc_link
*l
)
355 int tipc_link_prio(struct tipc_link
*l
)
360 unsigned long tipc_link_tolerance(struct tipc_link
*l
)
365 struct sk_buff_head
*tipc_link_inputq(struct tipc_link
*l
)
370 char tipc_link_plane(struct tipc_link
*l
)
375 struct net
*tipc_link_net(struct tipc_link
*l
)
380 void tipc_link_update_caps(struct tipc_link
*l
, u16 capabilities
)
382 l
->peer_caps
= capabilities
;
385 void tipc_link_add_bc_peer(struct tipc_link
*snd_l
,
386 struct tipc_link
*uc_l
,
387 struct sk_buff_head
*xmitq
)
389 struct tipc_link
*rcv_l
= uc_l
->bc_rcvlink
;
392 rcv_l
->acked
= snd_l
->snd_nxt
- 1;
393 snd_l
->state
= LINK_ESTABLISHED
;
394 tipc_link_build_bc_init_msg(uc_l
, xmitq
);
397 void tipc_link_remove_bc_peer(struct tipc_link
*snd_l
,
398 struct tipc_link
*rcv_l
,
399 struct sk_buff_head
*xmitq
)
401 u16 ack
= snd_l
->snd_nxt
- 1;
404 rcv_l
->bc_peer_is_up
= true;
405 rcv_l
->state
= LINK_ESTABLISHED
;
406 tipc_link_bc_ack_rcv(rcv_l
, ack
, 0, NULL
, xmitq
, NULL
);
407 trace_tipc_link_reset(rcv_l
, TIPC_DUMP_ALL
, "bclink removed!");
408 tipc_link_reset(rcv_l
);
409 rcv_l
->state
= LINK_RESET
;
410 if (!snd_l
->ackers
) {
411 trace_tipc_link_reset(snd_l
, TIPC_DUMP_ALL
, "zero ackers!");
412 tipc_link_reset(snd_l
);
413 snd_l
->state
= LINK_RESET
;
414 __skb_queue_purge(xmitq
);
418 int tipc_link_bc_peers(struct tipc_link
*l
)
423 static u16
link_bc_rcv_gap(struct tipc_link
*l
)
425 struct sk_buff
*skb
= skb_peek(&l
->deferdq
);
428 if (more(l
->snd_nxt
, l
->rcv_nxt
))
429 gap
= l
->snd_nxt
- l
->rcv_nxt
;
431 gap
= buf_seqno(skb
) - l
->rcv_nxt
;
435 void tipc_link_set_mtu(struct tipc_link
*l
, int mtu
)
440 int tipc_link_mtu(struct tipc_link
*l
)
445 int tipc_link_mss(struct tipc_link
*l
)
447 #ifdef CONFIG_TIPC_CRYPTO
448 return l
->mtu
- INT_H_SIZE
- EMSG_OVERHEAD
;
450 return l
->mtu
- INT_H_SIZE
;
454 u16
tipc_link_rcv_nxt(struct tipc_link
*l
)
459 u16
tipc_link_acked(struct tipc_link
*l
)
464 char *tipc_link_name(struct tipc_link
*l
)
469 u32
tipc_link_state(struct tipc_link
*l
)
475 * tipc_link_create - create a new link
476 * @net: pointer to associated network namespace
477 * @if_name: associated interface name
478 * @bearer_id: id (index) of associated bearer
479 * @tolerance: link tolerance to be used by link
480 * @net_plane: network plane (A,B,c..) this link belongs to
481 * @mtu: mtu to be advertised by link
482 * @priority: priority to be used by link
483 * @min_win: minimal send window to be used by link
484 * @max_win: maximal send window to be used by link
485 * @session: session to be used by link
486 * @peer: node id of peer node
487 * @peer_caps: bitmap describing peer node capabilities
488 * @bc_sndlink: the namespace global link used for broadcast sending
489 * @bc_rcvlink: the peer specific link used for broadcast reception
490 * @inputq: queue to put messages ready for delivery
491 * @namedq: queue to put binding table update messages ready for delivery
492 * @link: return value, pointer to put the created link
493 * @self: local unicast link id
494 * @peer_id: 128-bit ID of peer
496 * Return: true if link was created, otherwise false
498 bool tipc_link_create(struct net
*net
, char *if_name
, int bearer_id
,
499 int tolerance
, char net_plane
, u32 mtu
, int priority
,
500 u32 min_win
, u32 max_win
, u32 session
, u32 self
,
501 u32 peer
, u8
*peer_id
, u16 peer_caps
,
502 struct tipc_link
*bc_sndlink
,
503 struct tipc_link
*bc_rcvlink
,
504 struct sk_buff_head
*inputq
,
505 struct sk_buff_head
*namedq
,
506 struct tipc_link
**link
)
508 char peer_str
[NODE_ID_STR_LEN
] = {0,};
509 char self_str
[NODE_ID_STR_LEN
] = {0,};
512 l
= kzalloc(sizeof(*l
), GFP_ATOMIC
);
516 l
->session
= session
;
518 /* Set link name for unicast links only */
520 tipc_nodeid2string(self_str
, tipc_own_id(net
));
521 if (strlen(self_str
) > 16)
522 sprintf(self_str
, "%x", self
);
523 tipc_nodeid2string(peer_str
, peer_id
);
524 if (strlen(peer_str
) > 16)
525 sprintf(peer_str
, "%x", peer
);
527 /* Peer i/f name will be completed by reset/activate message */
528 snprintf(l
->name
, sizeof(l
->name
), "%s:%s-%s:unknown",
529 self_str
, if_name
, peer_str
);
531 strcpy(l
->if_name
, if_name
);
533 l
->peer_caps
= peer_caps
;
535 l
->in_session
= false;
536 l
->bearer_id
= bearer_id
;
537 l
->tolerance
= tolerance
;
539 bc_rcvlink
->tolerance
= tolerance
;
540 l
->net_plane
= net_plane
;
541 l
->advertised_mtu
= mtu
;
543 l
->priority
= priority
;
544 tipc_link_set_queue_limits(l
, min_win
, max_win
);
546 l
->bc_sndlink
= bc_sndlink
;
547 l
->bc_rcvlink
= bc_rcvlink
;
550 l
->state
= LINK_RESETTING
;
551 __skb_queue_head_init(&l
->transmq
);
552 __skb_queue_head_init(&l
->backlogq
);
553 __skb_queue_head_init(&l
->deferdq
);
554 __skb_queue_head_init(&l
->failover_deferdq
);
555 skb_queue_head_init(&l
->wakeupq
);
556 skb_queue_head_init(l
->inputq
);
561 * tipc_link_bc_create - create new link to be used for broadcast
562 * @net: pointer to associated network namespace
563 * @mtu: mtu to be used initially if no peers
564 * @min_win: minimal send window to be used by link
565 * @max_win: maximal send window to be used by link
566 * @inputq: queue to put messages ready for delivery
567 * @namedq: queue to put binding table update messages ready for delivery
568 * @link: return value, pointer to put the created link
569 * @ownnode: identity of own node
570 * @peer: node id of peer node
571 * @peer_id: 128-bit ID of peer
572 * @peer_caps: bitmap describing peer node capabilities
573 * @bc_sndlink: the namespace global link used for broadcast sending
575 * Return: true if link was created, otherwise false
577 bool tipc_link_bc_create(struct net
*net
, u32 ownnode
, u32 peer
, u8
*peer_id
,
578 int mtu
, u32 min_win
, u32 max_win
, u16 peer_caps
,
579 struct sk_buff_head
*inputq
,
580 struct sk_buff_head
*namedq
,
581 struct tipc_link
*bc_sndlink
,
582 struct tipc_link
**link
)
586 if (!tipc_link_create(net
, "", MAX_BEARERS
, 0, 'Z', mtu
, 0, min_win
,
587 max_win
, 0, ownnode
, peer
, NULL
, peer_caps
,
588 bc_sndlink
, NULL
, inputq
, namedq
, link
))
593 char peer_str
[NODE_ID_STR_LEN
] = {0,};
595 tipc_nodeid2string(peer_str
, peer_id
);
596 if (strlen(peer_str
) > 16)
597 sprintf(peer_str
, "%x", peer
);
598 /* Broadcast receiver link name: "broadcast-link:<peer>" */
599 snprintf(l
->name
, sizeof(l
->name
), "%s:%s", tipc_bclink_name
,
602 strcpy(l
->name
, tipc_bclink_name
);
604 trace_tipc_link_reset(l
, TIPC_DUMP_ALL
, "bclink created!");
606 l
->state
= LINK_RESET
;
610 /* Broadcast send link is always up */
611 if (link_is_bc_sndlink(l
))
612 l
->state
= LINK_ESTABLISHED
;
614 /* Disable replicast if even a single peer doesn't support it */
615 if (link_is_bc_rcvlink(l
) && !(peer_caps
& TIPC_BCAST_RCAST
))
616 tipc_bcast_toggle_rcast(net
, false);
622 * tipc_link_fsm_evt - link finite state machine
623 * @l: pointer to link
624 * @evt: state machine event to be processed
626 int tipc_link_fsm_evt(struct tipc_link
*l
, int evt
)
629 int old_state
= l
->state
;
634 case LINK_PEER_RESET_EVT
:
635 l
->state
= LINK_PEER_RESET
;
638 l
->state
= LINK_RESET
;
640 case LINK_FAILURE_EVT
:
641 case LINK_FAILOVER_BEGIN_EVT
:
642 case LINK_ESTABLISH_EVT
:
643 case LINK_FAILOVER_END_EVT
:
644 case LINK_SYNCH_BEGIN_EVT
:
645 case LINK_SYNCH_END_EVT
:
652 case LINK_PEER_RESET_EVT
:
653 l
->state
= LINK_ESTABLISHING
;
655 case LINK_FAILOVER_BEGIN_EVT
:
656 l
->state
= LINK_FAILINGOVER
;
658 case LINK_FAILURE_EVT
:
660 case LINK_ESTABLISH_EVT
:
661 case LINK_FAILOVER_END_EVT
:
663 case LINK_SYNCH_BEGIN_EVT
:
664 case LINK_SYNCH_END_EVT
:
669 case LINK_PEER_RESET
:
672 l
->state
= LINK_ESTABLISHING
;
674 case LINK_PEER_RESET_EVT
:
675 case LINK_ESTABLISH_EVT
:
676 case LINK_FAILURE_EVT
:
678 case LINK_SYNCH_BEGIN_EVT
:
679 case LINK_SYNCH_END_EVT
:
680 case LINK_FAILOVER_BEGIN_EVT
:
681 case LINK_FAILOVER_END_EVT
:
686 case LINK_FAILINGOVER
:
688 case LINK_FAILOVER_END_EVT
:
689 l
->state
= LINK_RESET
;
691 case LINK_PEER_RESET_EVT
:
693 case LINK_ESTABLISH_EVT
:
694 case LINK_FAILURE_EVT
:
696 case LINK_FAILOVER_BEGIN_EVT
:
697 case LINK_SYNCH_BEGIN_EVT
:
698 case LINK_SYNCH_END_EVT
:
703 case LINK_ESTABLISHING
:
705 case LINK_ESTABLISH_EVT
:
706 l
->state
= LINK_ESTABLISHED
;
708 case LINK_FAILOVER_BEGIN_EVT
:
709 l
->state
= LINK_FAILINGOVER
;
712 l
->state
= LINK_RESET
;
714 case LINK_FAILURE_EVT
:
715 case LINK_PEER_RESET_EVT
:
716 case LINK_SYNCH_BEGIN_EVT
:
717 case LINK_FAILOVER_END_EVT
:
719 case LINK_SYNCH_END_EVT
:
724 case LINK_ESTABLISHED
:
726 case LINK_PEER_RESET_EVT
:
727 l
->state
= LINK_PEER_RESET
;
728 rc
|= TIPC_LINK_DOWN_EVT
;
730 case LINK_FAILURE_EVT
:
731 l
->state
= LINK_RESETTING
;
732 rc
|= TIPC_LINK_DOWN_EVT
;
735 l
->state
= LINK_RESET
;
737 case LINK_ESTABLISH_EVT
:
738 case LINK_SYNCH_END_EVT
:
740 case LINK_SYNCH_BEGIN_EVT
:
741 l
->state
= LINK_SYNCHING
;
743 case LINK_FAILOVER_BEGIN_EVT
:
744 case LINK_FAILOVER_END_EVT
:
751 case LINK_PEER_RESET_EVT
:
752 l
->state
= LINK_PEER_RESET
;
753 rc
|= TIPC_LINK_DOWN_EVT
;
755 case LINK_FAILURE_EVT
:
756 l
->state
= LINK_RESETTING
;
757 rc
|= TIPC_LINK_DOWN_EVT
;
760 l
->state
= LINK_RESET
;
762 case LINK_ESTABLISH_EVT
:
763 case LINK_SYNCH_BEGIN_EVT
:
765 case LINK_SYNCH_END_EVT
:
766 l
->state
= LINK_ESTABLISHED
;
768 case LINK_FAILOVER_BEGIN_EVT
:
769 case LINK_FAILOVER_END_EVT
:
775 pr_err("Unknown FSM state %x in %s\n", l
->state
, l
->name
);
777 trace_tipc_link_fsm(l
->name
, old_state
, l
->state
, evt
);
780 pr_err("Illegal FSM event %x in state %x on link %s\n",
781 evt
, l
->state
, l
->name
);
782 trace_tipc_link_fsm(l
->name
, old_state
, l
->state
, evt
);
786 /* link_profile_stats - update statistical profiling of traffic
788 static void link_profile_stats(struct tipc_link
*l
)
791 struct tipc_msg
*msg
;
794 /* Update counters used in statistical profiling of send traffic */
795 l
->stats
.accu_queue_sz
+= skb_queue_len(&l
->transmq
);
796 l
->stats
.queue_sz_counts
++;
798 skb
= skb_peek(&l
->transmq
);
802 length
= msg_size(msg
);
804 if (msg_user(msg
) == MSG_FRAGMENTER
) {
805 if (msg_type(msg
) != FIRST_FRAGMENT
)
807 length
= msg_size(msg_inner_hdr(msg
));
809 l
->stats
.msg_lengths_total
+= length
;
810 l
->stats
.msg_length_counts
++;
812 l
->stats
.msg_length_profile
[0]++;
813 else if (length
<= 256)
814 l
->stats
.msg_length_profile
[1]++;
815 else if (length
<= 1024)
816 l
->stats
.msg_length_profile
[2]++;
817 else if (length
<= 4096)
818 l
->stats
.msg_length_profile
[3]++;
819 else if (length
<= 16384)
820 l
->stats
.msg_length_profile
[4]++;
821 else if (length
<= 32768)
822 l
->stats
.msg_length_profile
[5]++;
824 l
->stats
.msg_length_profile
[6]++;
828 * tipc_link_too_silent - check if link is "too silent"
829 * @l: tipc link to be checked
831 * Return: true if the link 'silent_intv_cnt' is about to reach the
832 * 'abort_limit' value, otherwise false
834 bool tipc_link_too_silent(struct tipc_link
*l
)
836 return (l
->silent_intv_cnt
+ 2 > l
->abort_limit
);
839 /* tipc_link_timeout - perform periodic task as instructed from node timeout
841 int tipc_link_timeout(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
848 u16 bc_snt
= l
->bc_sndlink
->snd_nxt
- 1;
849 u16 bc_acked
= l
->bc_rcvlink
->acked
;
850 struct tipc_mon_state
*mstate
= &l
->mon_state
;
852 trace_tipc_link_timeout(l
, TIPC_DUMP_NONE
, " ");
853 trace_tipc_link_too_silent(l
, TIPC_DUMP_ALL
, " ");
855 case LINK_ESTABLISHED
:
858 link_profile_stats(l
);
859 tipc_mon_get_state(l
->net
, l
->addr
, mstate
, l
->bearer_id
);
860 if (mstate
->reset
|| (l
->silent_intv_cnt
> l
->abort_limit
))
861 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
862 state
= bc_acked
!= bc_snt
;
863 state
|= l
->bc_rcvlink
->rcv_unacked
;
864 state
|= l
->rcv_unacked
;
865 state
|= !skb_queue_empty(&l
->transmq
);
866 probe
= mstate
->probing
;
867 probe
|= l
->silent_intv_cnt
;
868 if (probe
|| mstate
->monitoring
)
869 l
->silent_intv_cnt
++;
870 probe
|= !skb_queue_empty(&l
->deferdq
);
871 if (l
->snd_nxt
== l
->checkpoint
) {
872 tipc_link_update_cwin(l
, 0, 0);
875 l
->checkpoint
= l
->snd_nxt
;
878 setup
= l
->rst_cnt
++ <= 4;
879 setup
|= !(l
->rst_cnt
% 16);
882 case LINK_ESTABLISHING
:
886 case LINK_PEER_RESET
:
888 case LINK_FAILINGOVER
:
894 if (state
|| probe
|| setup
)
895 tipc_link_build_proto_msg(l
, mtyp
, probe
, 0, 0, 0, 0, xmitq
);
901 * link_schedule_user - schedule a message sender for wakeup after congestion
903 * @hdr: header of message that is being sent
904 * Create pseudo msg to send back to user when congestion abates
906 static int link_schedule_user(struct tipc_link
*l
, struct tipc_msg
*hdr
)
908 u32 dnode
= tipc_own_addr(l
->net
);
909 u32 dport
= msg_origport(hdr
);
912 /* Create and schedule wakeup pseudo message */
913 skb
= tipc_msg_create(SOCK_WAKEUP
, 0, INT_H_SIZE
, 0,
914 dnode
, l
->addr
, dport
, 0, 0);
917 msg_set_dest_droppable(buf_msg(skb
), true);
918 TIPC_SKB_CB(skb
)->chain_imp
= msg_importance(hdr
);
919 skb_queue_tail(&l
->wakeupq
, skb
);
920 l
->stats
.link_congs
++;
921 trace_tipc_link_conges(l
, TIPC_DUMP_ALL
, "wakeup scheduled!");
926 * link_prepare_wakeup - prepare users for wakeup after congestion
928 * Wake up a number of waiting users, as permitted by available space
931 static void link_prepare_wakeup(struct tipc_link
*l
)
933 struct sk_buff_head
*wakeupq
= &l
->wakeupq
;
934 struct sk_buff_head
*inputq
= l
->inputq
;
935 struct sk_buff
*skb
, *tmp
;
936 struct sk_buff_head tmpq
;
940 __skb_queue_head_init(&tmpq
);
942 for (; imp
<= TIPC_SYSTEM_IMPORTANCE
; imp
++)
943 avail
[imp
] = l
->backlog
[imp
].limit
- l
->backlog
[imp
].len
;
945 skb_queue_walk_safe(wakeupq
, skb
, tmp
) {
946 imp
= TIPC_SKB_CB(skb
)->chain_imp
;
950 __skb_unlink(skb
, wakeupq
);
951 __skb_queue_tail(&tmpq
, skb
);
954 spin_lock_bh(&inputq
->lock
);
955 skb_queue_splice_tail(&tmpq
, inputq
);
956 spin_unlock_bh(&inputq
->lock
);
961 * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
962 * the given skb should be next attempted
963 * @skb: skb to set a future retransmission time for
964 * @l: link the skb will be transmitted on
966 static void tipc_link_set_skb_retransmit_time(struct sk_buff
*skb
,
969 if (link_is_bc_sndlink(l
))
970 TIPC_SKB_CB(skb
)->nxt_retr
= TIPC_BC_RETR_LIM
;
972 TIPC_SKB_CB(skb
)->nxt_retr
= TIPC_UC_RETR_TIME
;
975 void tipc_link_reset(struct tipc_link
*l
)
977 struct sk_buff_head list
;
980 __skb_queue_head_init(&list
);
982 l
->in_session
= false;
983 /* Force re-synch of peer session number before establishing */
986 l
->mtu
= l
->advertised_mtu
;
988 spin_lock_bh(&l
->wakeupq
.lock
);
989 skb_queue_splice_init(&l
->wakeupq
, &list
);
990 spin_unlock_bh(&l
->wakeupq
.lock
);
992 spin_lock_bh(&l
->inputq
->lock
);
993 skb_queue_splice_init(&list
, l
->inputq
);
994 spin_unlock_bh(&l
->inputq
->lock
);
996 __skb_queue_purge(&l
->transmq
);
997 __skb_queue_purge(&l
->deferdq
);
998 __skb_queue_purge(&l
->backlogq
);
999 __skb_queue_purge(&l
->failover_deferdq
);
1000 for (imp
= 0; imp
<= TIPC_SYSTEM_IMPORTANCE
; imp
++) {
1001 l
->backlog
[imp
].len
= 0;
1002 l
->backlog
[imp
].target_bskb
= NULL
;
1004 kfree_skb(l
->reasm_buf
);
1005 kfree_skb(l
->reasm_tnlmsg
);
1006 kfree_skb(l
->failover_reasm_skb
);
1007 l
->reasm_buf
= NULL
;
1008 l
->reasm_tnlmsg
= NULL
;
1009 l
->failover_reasm_skb
= NULL
;
1013 l
->snd_nxt_state
= 1;
1014 l
->rcv_nxt_state
= 1;
1019 l
->silent_intv_cnt
= 0;
1021 l
->bc_peer_is_up
= false;
1022 memset(&l
->mon_state
, 0, sizeof(l
->mon_state
));
1023 tipc_link_reset_stats(l
);
1027 * tipc_link_xmit(): enqueue buffer list according to queue situation
1029 * @list: chain of buffers containing message
1030 * @xmitq: returned list of packets to be sent by caller
1032 * Consumes the buffer chain.
1033 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
1034 * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
1036 int tipc_link_xmit(struct tipc_link
*l
, struct sk_buff_head
*list
,
1037 struct sk_buff_head
*xmitq
)
1039 struct sk_buff_head
*backlogq
= &l
->backlogq
;
1040 struct sk_buff_head
*transmq
= &l
->transmq
;
1041 struct sk_buff
*skb
, *_skb
;
1042 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
1043 u16 ack
= l
->rcv_nxt
- 1;
1044 u16 seqno
= l
->snd_nxt
;
1045 int pkt_cnt
= skb_queue_len(list
);
1046 unsigned int mss
= tipc_link_mss(l
);
1047 unsigned int cwin
= l
->window
;
1048 unsigned int mtu
= l
->mtu
;
1049 struct tipc_msg
*hdr
;
1057 hdr
= buf_msg(skb_peek(list
));
1058 if (unlikely(msg_size(hdr
) > mtu
)) {
1059 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1060 skb_queue_len(list
), msg_user(hdr
),
1061 msg_type(hdr
), msg_size(hdr
), mtu
);
1062 __skb_queue_purge(list
);
1066 imp
= msg_importance(hdr
);
1067 /* Allow oversubscription of one data msg per source at congestion */
1068 if (unlikely(l
->backlog
[imp
].len
>= l
->backlog
[imp
].limit
)) {
1069 if (imp
== TIPC_SYSTEM_IMPORTANCE
) {
1070 pr_warn("%s<%s>, link overflow", link_rst_msg
, l
->name
);
1073 rc
= link_schedule_user(l
, hdr
);
1077 l
->stats
.sent_fragmented
++;
1078 l
->stats
.sent_fragments
+= pkt_cnt
;
1081 /* Prepare each packet for sending, and add to relevant queue: */
1082 while ((skb
= __skb_dequeue(list
))) {
1083 if (likely(skb_queue_len(transmq
) < cwin
)) {
1085 msg_set_seqno(hdr
, seqno
);
1086 msg_set_ack(hdr
, ack
);
1087 msg_set_bcast_ack(hdr
, bc_ack
);
1088 _skb
= skb_clone(skb
, GFP_ATOMIC
);
1091 __skb_queue_purge(list
);
1094 __skb_queue_tail(transmq
, skb
);
1095 tipc_link_set_skb_retransmit_time(skb
, l
);
1096 __skb_queue_tail(xmitq
, _skb
);
1097 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
1099 l
->stats
.sent_pkts
++;
1103 if (tipc_msg_try_bundle(l
->backlog
[imp
].target_bskb
, &skb
,
1104 mss
, l
->addr
, &new_bundle
)) {
1106 /* Keep a ref. to the skb for next try */
1107 l
->backlog
[imp
].target_bskb
= skb
;
1108 l
->backlog
[imp
].len
++;
1109 __skb_queue_tail(backlogq
, skb
);
1112 l
->stats
.sent_bundles
++;
1113 l
->stats
.sent_bundled
++;
1115 l
->stats
.sent_bundled
++;
1119 l
->backlog
[imp
].target_bskb
= NULL
;
1120 l
->backlog
[imp
].len
+= (1 + skb_queue_len(list
));
1121 __skb_queue_tail(backlogq
, skb
);
1122 skb_queue_splice_tail_init(list
, backlogq
);
1128 static void tipc_link_update_cwin(struct tipc_link
*l
, int released
,
1131 int bklog_len
= skb_queue_len(&l
->backlogq
);
1132 struct sk_buff_head
*txq
= &l
->transmq
;
1133 int txq_len
= skb_queue_len(txq
);
1134 u16 cwin
= l
->window
;
1136 /* Enter fast recovery */
1137 if (unlikely(retransmitted
)) {
1138 l
->ssthresh
= max_t(u16
, l
->window
/ 2, 300);
1139 l
->window
= min_t(u16
, l
->ssthresh
, l
->window
);
1142 /* Enter slow start */
1143 if (unlikely(!released
)) {
1144 l
->ssthresh
= max_t(u16
, l
->window
/ 2, 300);
1145 l
->window
= l
->min_win
;
1148 /* Don't increase window if no pressure on the transmit queue */
1149 if (txq_len
+ bklog_len
< cwin
)
1152 /* Don't increase window if there are holes the transmit queue */
1153 if (txq_len
&& l
->snd_nxt
- buf_seqno(skb_peek(txq
)) != txq_len
)
1156 l
->cong_acks
+= released
;
1159 if (cwin
<= l
->ssthresh
) {
1160 l
->window
= min_t(u16
, cwin
+ released
, l
->max_win
);
1163 /* Congestion avoidance */
1164 if (l
->cong_acks
< cwin
)
1166 l
->window
= min_t(u16
, ++cwin
, l
->max_win
);
1170 static void tipc_link_advance_backlog(struct tipc_link
*l
,
1171 struct sk_buff_head
*xmitq
)
1173 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
1174 struct sk_buff_head
*txq
= &l
->transmq
;
1175 struct sk_buff
*skb
, *_skb
;
1176 u16 ack
= l
->rcv_nxt
- 1;
1177 u16 seqno
= l
->snd_nxt
;
1178 struct tipc_msg
*hdr
;
1179 u16 cwin
= l
->window
;
1182 while (skb_queue_len(txq
) < cwin
) {
1183 skb
= skb_peek(&l
->backlogq
);
1186 _skb
= skb_clone(skb
, GFP_ATOMIC
);
1189 __skb_dequeue(&l
->backlogq
);
1191 imp
= msg_importance(hdr
);
1192 l
->backlog
[imp
].len
--;
1193 if (unlikely(skb
== l
->backlog
[imp
].target_bskb
))
1194 l
->backlog
[imp
].target_bskb
= NULL
;
1195 __skb_queue_tail(&l
->transmq
, skb
);
1196 tipc_link_set_skb_retransmit_time(skb
, l
);
1198 __skb_queue_tail(xmitq
, _skb
);
1199 TIPC_SKB_CB(skb
)->ackers
= l
->ackers
;
1200 msg_set_seqno(hdr
, seqno
);
1201 msg_set_ack(hdr
, ack
);
1202 msg_set_bcast_ack(hdr
, bc_ack
);
1204 l
->stats
.sent_pkts
++;
1211 * link_retransmit_failure() - Detect repeated retransmit failures
1212 * @l: tipc link sender
1213 * @r: tipc link receiver (= l in case of unicast)
1214 * @rc: returned code
1216 * Return: true if the repeated retransmit failures happens, otherwise
1219 static bool link_retransmit_failure(struct tipc_link
*l
, struct tipc_link
*r
,
1222 struct sk_buff
*skb
= skb_peek(&l
->transmq
);
1223 struct tipc_msg
*hdr
;
1228 if (!TIPC_SKB_CB(skb
)->retr_cnt
)
1231 if (!time_after(jiffies
, TIPC_SKB_CB(skb
)->retr_stamp
+
1232 msecs_to_jiffies(r
->tolerance
* 10)))
1236 if (link_is_bc_sndlink(l
) && !less(r
->acked
, msg_seqno(hdr
)))
1239 pr_warn("Retransmission failure on link <%s>\n", l
->name
);
1240 link_print(l
, "State of link ");
1241 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1242 msg_user(hdr
), msg_type(hdr
), msg_size(hdr
), msg_errcode(hdr
));
1243 pr_info("sqno %u, prev: %x, dest: %x\n",
1244 msg_seqno(hdr
), msg_prevnode(hdr
), msg_destnode(hdr
));
1245 pr_info("retr_stamp %d, retr_cnt %d\n",
1246 jiffies_to_msecs(TIPC_SKB_CB(skb
)->retr_stamp
),
1247 TIPC_SKB_CB(skb
)->retr_cnt
);
1249 trace_tipc_list_dump(&l
->transmq
, true, "retrans failure!");
1250 trace_tipc_link_dump(l
, TIPC_DUMP_NONE
, "retrans failure!");
1251 trace_tipc_link_dump(r
, TIPC_DUMP_NONE
, "retrans failure!");
1253 if (link_is_bc_sndlink(l
)) {
1254 r
->state
= LINK_RESET
;
1255 *rc
|= TIPC_LINK_DOWN_EVT
;
1257 *rc
|= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1263 /* tipc_data_input - deliver data and name distr msgs to upper layer
1265 * Consumes buffer if message is of right type
1266 * Node lock must be held
1268 static bool tipc_data_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1269 struct sk_buff_head
*inputq
)
1271 struct sk_buff_head
*mc_inputq
= l
->bc_rcvlink
->inputq
;
1272 struct tipc_msg
*hdr
= buf_msg(skb
);
1274 switch (msg_user(hdr
)) {
1275 case TIPC_LOW_IMPORTANCE
:
1276 case TIPC_MEDIUM_IMPORTANCE
:
1277 case TIPC_HIGH_IMPORTANCE
:
1278 case TIPC_CRITICAL_IMPORTANCE
:
1279 if (unlikely(msg_in_group(hdr
) || msg_mcast(hdr
))) {
1280 skb_queue_tail(mc_inputq
, skb
);
1285 skb_queue_tail(inputq
, skb
);
1287 case GROUP_PROTOCOL
:
1288 skb_queue_tail(mc_inputq
, skb
);
1290 case NAME_DISTRIBUTOR
:
1291 l
->bc_rcvlink
->state
= LINK_ESTABLISHED
;
1292 skb_queue_tail(l
->namedq
, skb
);
1295 case TUNNEL_PROTOCOL
:
1296 case MSG_FRAGMENTER
:
1297 case BCAST_PROTOCOL
:
1299 #ifdef CONFIG_TIPC_CRYPTO
1301 if (TIPC_SKB_CB(skb
)->decrypted
) {
1302 tipc_crypto_msg_rcv(l
->net
, skb
);
1308 pr_warn("Dropping received illegal msg type\n");
1314 /* tipc_link_input - process packet that has passed link protocol check
1318 static int tipc_link_input(struct tipc_link
*l
, struct sk_buff
*skb
,
1319 struct sk_buff_head
*inputq
,
1320 struct sk_buff
**reasm_skb
)
1322 struct tipc_msg
*hdr
= buf_msg(skb
);
1323 struct sk_buff
*iskb
;
1324 struct sk_buff_head tmpq
;
1325 int usr
= msg_user(hdr
);
1328 if (usr
== MSG_BUNDLER
) {
1329 skb_queue_head_init(&tmpq
);
1330 l
->stats
.recv_bundles
++;
1331 l
->stats
.recv_bundled
+= msg_msgcnt(hdr
);
1332 while (tipc_msg_extract(skb
, &iskb
, &pos
))
1333 tipc_data_input(l
, iskb
, &tmpq
);
1334 tipc_skb_queue_splice_tail(&tmpq
, inputq
);
1336 } else if (usr
== MSG_FRAGMENTER
) {
1337 l
->stats
.recv_fragments
++;
1338 if (tipc_buf_append(reasm_skb
, &skb
)) {
1339 l
->stats
.recv_fragmented
++;
1340 tipc_data_input(l
, skb
, inputq
);
1341 } else if (!*reasm_skb
&& !link_is_bc_rcvlink(l
)) {
1342 pr_warn_ratelimited("Unable to build fragment list\n");
1343 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1346 } else if (usr
== BCAST_PROTOCOL
) {
1347 tipc_bcast_lock(l
->net
);
1348 tipc_link_bc_init_rcv(l
->bc_rcvlink
, hdr
);
1349 tipc_bcast_unlock(l
->net
);
1356 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1357 * inner message along with the ones in the old link's
1360 * @skb: TUNNEL_PROTOCOL message
1361 * @inputq: queue to put messages ready for delivery
1363 static int tipc_link_tnl_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1364 struct sk_buff_head
*inputq
)
1366 struct sk_buff
**reasm_skb
= &l
->failover_reasm_skb
;
1367 struct sk_buff
**reasm_tnlmsg
= &l
->reasm_tnlmsg
;
1368 struct sk_buff_head
*fdefq
= &l
->failover_deferdq
;
1369 struct tipc_msg
*hdr
= buf_msg(skb
);
1370 struct sk_buff
*iskb
;
1375 if (msg_type(hdr
) == SYNCH_MSG
) {
1380 /* Not a fragment? */
1381 if (likely(!msg_nof_fragms(hdr
))) {
1382 if (unlikely(!tipc_msg_extract(skb
, &iskb
, &ipos
))) {
1383 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1384 skb_queue_len(fdefq
));
1389 /* Set fragment type for buf_append */
1390 if (msg_fragm_no(hdr
) == 1)
1391 msg_set_type(hdr
, FIRST_FRAGMENT
);
1392 else if (msg_fragm_no(hdr
) < msg_nof_fragms(hdr
))
1393 msg_set_type(hdr
, FRAGMENT
);
1395 msg_set_type(hdr
, LAST_FRAGMENT
);
1397 if (!tipc_buf_append(reasm_tnlmsg
, &skb
)) {
1398 /* Successful but non-complete reassembly? */
1399 if (*reasm_tnlmsg
|| link_is_bc_rcvlink(l
))
1401 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1402 return tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
1408 seqno
= buf_seqno(iskb
);
1409 if (unlikely(less(seqno
, l
->drop_point
))) {
1413 if (unlikely(seqno
!= l
->drop_point
)) {
1414 __tipc_skb_queue_sorted(fdefq
, seqno
, iskb
);
1419 if (!tipc_data_input(l
, iskb
, inputq
))
1420 rc
|= tipc_link_input(l
, iskb
, inputq
, reasm_skb
);
1423 } while ((iskb
= __tipc_skb_dequeue(fdefq
, l
->drop_point
)));
1429 * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1430 * @ga: returned pointer to the Gap ACK blocks if any
1432 * @hdr: the PROTOCOL/STATE_MSG header
1433 * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1435 * Return: the total Gap ACK blocks size
1437 u16
tipc_get_gap_ack_blks(struct tipc_gap_ack_blks
**ga
, struct tipc_link
*l
,
1438 struct tipc_msg
*hdr
, bool uc
)
1440 struct tipc_gap_ack_blks
*p
;
1443 /* Does peer support the Gap ACK blocks feature? */
1444 if (l
->peer_caps
& TIPC_GAP_ACK_BLOCK
) {
1445 p
= (struct tipc_gap_ack_blks
*)msg_data(hdr
);
1448 if (sz
== struct_size(p
, gacks
, p
->ugack_cnt
+ p
->bgack_cnt
)) {
1449 /* Good, check if the desired type exists */
1450 if ((uc
&& p
->ugack_cnt
) || (!uc
&& p
->bgack_cnt
))
1452 /* Backward compatible: peer might not support bc, but uc? */
1453 } else if (uc
&& sz
== struct_size(p
, gacks
, p
->ugack_cnt
)) {
1460 /* Other cases: ignore! */
1468 static u8
__tipc_build_gap_ack_blks(struct tipc_gap_ack_blks
*ga
,
1469 struct tipc_link
*l
, u8 start_index
)
1471 struct tipc_gap_ack
*gacks
= &ga
->gacks
[start_index
];
1472 struct sk_buff
*skb
= skb_peek(&l
->deferdq
);
1473 u16 expect
, seqno
= 0;
1479 expect
= buf_seqno(skb
);
1480 skb_queue_walk(&l
->deferdq
, skb
) {
1481 seqno
= buf_seqno(skb
);
1482 if (unlikely(more(seqno
, expect
))) {
1483 gacks
[n
].ack
= htons(expect
- 1);
1484 gacks
[n
].gap
= htons(seqno
- expect
);
1485 if (++n
>= MAX_GAP_ACK_BLKS
/ 2) {
1486 pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1488 skb_queue_len(&l
->deferdq
));
1491 } else if (unlikely(less(seqno
, expect
))) {
1492 pr_warn("Unexpected skb in deferdq!\n");
1499 gacks
[n
].ack
= htons(seqno
);
1505 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1506 * @l: tipc unicast link
1507 * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1509 * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1510 * links of a certain peer, the buffer after built has the network data format
1511 * as found at the struct tipc_gap_ack_blks definition.
1513 * returns the actual allocated memory size
1515 static u16
tipc_build_gap_ack_blks(struct tipc_link
*l
, struct tipc_msg
*hdr
)
1517 struct tipc_link
*bcl
= l
->bc_rcvlink
;
1518 struct tipc_gap_ack_blks
*ga
;
1521 ga
= (struct tipc_gap_ack_blks
*)msg_data(hdr
);
1523 /* Start with broadcast link first */
1524 tipc_bcast_lock(bcl
->net
);
1525 msg_set_bcast_ack(hdr
, bcl
->rcv_nxt
- 1);
1526 msg_set_bc_gap(hdr
, link_bc_rcv_gap(bcl
));
1527 ga
->bgack_cnt
= __tipc_build_gap_ack_blks(ga
, bcl
, 0);
1528 tipc_bcast_unlock(bcl
->net
);
1530 /* Now for unicast link, but an explicit NACK only (???) */
1531 ga
->ugack_cnt
= (msg_seq_gap(hdr
)) ?
1532 __tipc_build_gap_ack_blks(ga
, l
, ga
->bgack_cnt
) : 0;
1535 len
= struct_size(ga
, gacks
, ga
->bgack_cnt
+ ga
->ugack_cnt
);
1536 ga
->len
= htons(len
);
1540 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1541 * acked packets, also doing retransmissions if
1543 * @l: tipc link with transmq queue to be advanced
1544 * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1545 * @acked: seqno of last packet acked by peer without any gaps before
1546 * @gap: # of gap packets
1547 * @ga: buffer pointer to Gap ACK blocks from peer
1548 * @xmitq: queue for accumulating the retransmitted packets if any
1549 * @retransmitted: returned boolean value if a retransmission is really issued
1550 * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1551 * happens (- unlikely case)
1553 * Return: the number of packets released from the link transmq
1555 static int tipc_link_advance_transmq(struct tipc_link
*l
, struct tipc_link
*r
,
1557 struct tipc_gap_ack_blks
*ga
,
1558 struct sk_buff_head
*xmitq
,
1559 bool *retransmitted
, int *rc
)
1561 struct tipc_gap_ack_blks
*last_ga
= r
->last_ga
, *this_ga
= NULL
;
1562 struct tipc_gap_ack
*gacks
= NULL
;
1563 struct sk_buff
*skb
, *_skb
, *tmp
;
1564 struct tipc_msg
*hdr
;
1565 u32 qlen
= skb_queue_len(&l
->transmq
);
1566 u16 nacked
= acked
, ngap
= gap
, gack_cnt
= 0;
1567 u16 bc_ack
= l
->bc_rcvlink
->rcv_nxt
- 1;
1568 u16 ack
= l
->rcv_nxt
- 1;
1570 u16 end
= r
->acked
, start
= end
, offset
= r
->last_gap
;
1571 u16 si
= (last_ga
) ? last_ga
->start_index
: 0;
1572 bool is_uc
= !link_is_bc_sndlink(l
);
1573 bool bc_has_acked
= false;
1575 trace_tipc_link_retrans(r
, acked
+ 1, acked
+ gap
, &l
->transmq
);
1577 /* Determine Gap ACK blocks if any for the particular link */
1579 /* Get the Gap ACKs, uc part */
1580 gack_cnt
= ga
->ugack_cnt
;
1581 gacks
= &ga
->gacks
[ga
->bgack_cnt
];
1583 /* Copy the Gap ACKs, bc part, for later renewal if needed */
1584 this_ga
= kmemdup(ga
, struct_size(ga
, gacks
, ga
->bgack_cnt
),
1586 if (likely(this_ga
)) {
1587 this_ga
->start_index
= 0;
1588 /* Start with the bc Gap ACKs */
1589 gack_cnt
= this_ga
->bgack_cnt
;
1590 gacks
= &this_ga
->gacks
[0];
1592 /* Hmm, we can get in trouble..., simply ignore it */
1593 pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1597 /* Advance the link transmq */
1598 skb_queue_walk_safe(&l
->transmq
, skb
, tmp
) {
1599 seqno
= buf_seqno(skb
);
1602 if (less_eq(seqno
, nacked
)) {
1605 /* Skip packets peer has already acked */
1606 if (!more(seqno
, r
->acked
))
1608 /* Get the next of last Gap ACK blocks */
1609 while (more(seqno
, end
)) {
1610 if (!last_ga
|| si
>= last_ga
->bgack_cnt
)
1612 start
= end
+ offset
+ 1;
1613 end
= ntohs(last_ga
->gacks
[si
].ack
);
1614 offset
= ntohs(last_ga
->gacks
[si
].gap
);
1616 WARN_ONCE(more(start
, end
) ||
1618 si
< last_ga
->bgack_cnt
) ||
1619 si
> MAX_GAP_ACK_BLKS
,
1620 "Corrupted Gap ACK: %d %d %d %d %d\n",
1621 start
, end
, offset
, si
,
1622 last_ga
->bgack_cnt
);
1624 /* Check against the last Gap ACK block */
1625 if (in_range(seqno
, start
, end
))
1627 /* Update/release the packet peer is acking */
1628 bc_has_acked
= true;
1629 if (--TIPC_SKB_CB(skb
)->ackers
)
1633 __skb_unlink(skb
, &l
->transmq
);
1635 } else if (less_eq(seqno
, nacked
+ ngap
)) {
1636 /* First gap: check if repeated retrans failures? */
1637 if (unlikely(seqno
== acked
+ 1 &&
1638 link_retransmit_failure(l
, r
, rc
))) {
1639 /* Ignore this bc Gap ACKs if any */
1644 /* retransmit skb if unrestricted*/
1645 if (time_before(jiffies
, TIPC_SKB_CB(skb
)->nxt_retr
))
1647 tipc_link_set_skb_retransmit_time(skb
, l
);
1648 _skb
= pskb_copy(skb
, GFP_ATOMIC
);
1651 hdr
= buf_msg(_skb
);
1652 msg_set_ack(hdr
, ack
);
1653 msg_set_bcast_ack(hdr
, bc_ack
);
1654 _skb
->priority
= TC_PRIO_CONTROL
;
1655 __skb_queue_tail(xmitq
, _skb
);
1656 l
->stats
.retransmitted
++;
1658 r
->stats
.retransmitted
++;
1659 *retransmitted
= true;
1660 /* Increase actual retrans counter & mark first time */
1661 if (!TIPC_SKB_CB(skb
)->retr_cnt
++)
1662 TIPC_SKB_CB(skb
)->retr_stamp
= jiffies
;
1664 /* retry with Gap ACK blocks if any */
1667 nacked
= ntohs(gacks
[n
].ack
);
1668 ngap
= ntohs(gacks
[n
].gap
);
1674 /* Renew last Gap ACK blocks for bc if needed */
1678 r
->last_ga
= this_ga
;
1680 } else if (last_ga
) {
1681 if (less(acked
, start
)) {
1683 offset
= start
- acked
- 1;
1684 } else if (less(acked
, end
)) {
1687 if (si
< last_ga
->bgack_cnt
) {
1688 last_ga
->start_index
= si
;
1689 r
->last_gap
= offset
;
1703 return qlen
- skb_queue_len(&l
->transmq
);
1706 /* tipc_link_build_state_msg: prepare link state message for transmission
1708 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1709 * risk of ack storms towards the sender
1711 int tipc_link_build_state_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1716 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1717 if (link_is_bc_rcvlink(l
)) {
1718 if (((l
->rcv_nxt
^ tipc_own_addr(l
->net
)) & 0xf) != 0xf)
1722 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1723 l
->snd_nxt
= l
->rcv_nxt
;
1724 return TIPC_LINK_SND_STATE
;
1728 l
->stats
.sent_acks
++;
1729 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, 0, xmitq
);
1733 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1735 void tipc_link_build_reset_msg(struct tipc_link
*l
, struct sk_buff_head
*xmitq
)
1737 int mtyp
= RESET_MSG
;
1738 struct sk_buff
*skb
;
1740 if (l
->state
== LINK_ESTABLISHING
)
1741 mtyp
= ACTIVATE_MSG
;
1743 tipc_link_build_proto_msg(l
, mtyp
, 0, 0, 0, 0, 0, xmitq
);
1745 /* Inform peer that this endpoint is going down if applicable */
1746 skb
= skb_peek_tail(xmitq
);
1747 if (skb
&& (l
->state
== LINK_RESET
))
1748 msg_set_peer_stopping(buf_msg(skb
), 1);
1751 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1752 * Note that sending of broadcast NACK is coordinated among nodes, to
1753 * reduce the risk of NACK storms towards the sender
1755 static int tipc_link_build_nack_msg(struct tipc_link
*l
,
1756 struct sk_buff_head
*xmitq
)
1758 u32 def_cnt
= ++l
->stats
.deferred_recv
;
1759 struct sk_buff_head
*dfq
= &l
->deferdq
;
1760 u32 defq_len
= skb_queue_len(dfq
);
1763 if (link_is_bc_rcvlink(l
)) {
1764 match1
= def_cnt
& 0xf;
1765 match2
= tipc_own_addr(l
->net
) & 0xf;
1766 if (match1
== match2
)
1767 return TIPC_LINK_SND_STATE
;
1771 if (defq_len
>= 3 && !((defq_len
- 3) % 16)) {
1772 u16 rcvgap
= buf_seqno(skb_peek(dfq
)) - l
->rcv_nxt
;
1774 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0,
1775 rcvgap
, 0, 0, xmitq
);
1780 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1781 * @l: the link that should handle the message
1783 * @xmitq: queue to place packets to be sent after this call
1785 int tipc_link_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
1786 struct sk_buff_head
*xmitq
)
1788 struct sk_buff_head
*defq
= &l
->deferdq
;
1789 struct tipc_msg
*hdr
= buf_msg(skb
);
1790 u16 seqno
, rcv_nxt
, win_lim
;
1794 /* Verify and update link state */
1795 if (unlikely(msg_user(hdr
) == LINK_PROTOCOL
))
1796 return tipc_link_proto_rcv(l
, skb
, xmitq
);
1798 /* Don't send probe at next timeout expiration */
1799 l
->silent_intv_cnt
= 0;
1803 seqno
= msg_seqno(hdr
);
1804 rcv_nxt
= l
->rcv_nxt
;
1805 win_lim
= rcv_nxt
+ TIPC_MAX_LINK_WIN
;
1807 if (unlikely(!link_is_up(l
))) {
1808 if (l
->state
== LINK_ESTABLISHING
)
1809 rc
= TIPC_LINK_UP_EVT
;
1814 /* Drop if outside receive window */
1815 if (unlikely(less(seqno
, rcv_nxt
) || more(seqno
, win_lim
))) {
1816 l
->stats
.duplicates
++;
1820 released
+= tipc_link_advance_transmq(l
, l
, msg_ack(hdr
), 0,
1821 NULL
, NULL
, NULL
, NULL
);
1823 /* Defer delivery if sequence gap */
1824 if (unlikely(seqno
!= rcv_nxt
)) {
1825 if (!__tipc_skb_queue_sorted(defq
, seqno
, skb
))
1826 l
->stats
.duplicates
++;
1827 rc
|= tipc_link_build_nack_msg(l
, xmitq
);
1831 /* Deliver packet */
1833 l
->stats
.recv_pkts
++;
1835 if (unlikely(msg_user(hdr
) == TUNNEL_PROTOCOL
))
1836 rc
|= tipc_link_tnl_rcv(l
, skb
, l
->inputq
);
1837 else if (!tipc_data_input(l
, skb
, l
->inputq
))
1838 rc
|= tipc_link_input(l
, skb
, l
->inputq
, &l
->reasm_buf
);
1839 if (unlikely(++l
->rcv_unacked
>= TIPC_MIN_LINK_WIN
))
1840 rc
|= tipc_link_build_state_msg(l
, xmitq
);
1841 if (unlikely(rc
& ~TIPC_LINK_SND_STATE
))
1843 } while ((skb
= __tipc_skb_dequeue(defq
, l
->rcv_nxt
)));
1845 /* Forward queues and wake up waiting users */
1847 tipc_link_update_cwin(l
, released
, 0);
1848 tipc_link_advance_backlog(l
, xmitq
);
1849 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
1850 link_prepare_wakeup(l
);
1855 static void tipc_link_build_proto_msg(struct tipc_link
*l
, int mtyp
, bool probe
,
1856 bool probe_reply
, u16 rcvgap
,
1857 int tolerance
, int priority
,
1858 struct sk_buff_head
*xmitq
)
1860 struct tipc_mon_state
*mstate
= &l
->mon_state
;
1861 struct sk_buff_head
*dfq
= &l
->deferdq
;
1862 struct tipc_link
*bcl
= l
->bc_rcvlink
;
1863 struct tipc_msg
*hdr
;
1864 struct sk_buff
*skb
;
1865 bool node_up
= link_is_up(bcl
);
1866 u16 glen
= 0, bc_rcvgap
= 0;
1870 /* Don't send protocol message during reset or link failover */
1871 if (tipc_link_is_blocked(l
))
1874 if (!tipc_link_is_up(l
) && (mtyp
== STATE_MSG
))
1877 if ((probe
|| probe_reply
) && !skb_queue_empty(dfq
))
1878 rcvgap
= buf_seqno(skb_peek(dfq
)) - l
->rcv_nxt
;
1880 skb
= tipc_msg_create(LINK_PROTOCOL
, mtyp
, INT_H_SIZE
,
1881 tipc_max_domain_size
+ MAX_GAP_ACK_BLKS_SZ
,
1882 l
->addr
, tipc_own_addr(l
->net
), 0, 0, 0);
1887 data
= msg_data(hdr
);
1888 msg_set_session(hdr
, l
->session
);
1889 msg_set_bearer_id(hdr
, l
->bearer_id
);
1890 msg_set_net_plane(hdr
, l
->net_plane
);
1891 msg_set_next_sent(hdr
, l
->snd_nxt
);
1892 msg_set_ack(hdr
, l
->rcv_nxt
- 1);
1893 msg_set_bcast_ack(hdr
, bcl
->rcv_nxt
- 1);
1894 msg_set_bc_ack_invalid(hdr
, !node_up
);
1895 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
1896 msg_set_link_tolerance(hdr
, tolerance
);
1897 msg_set_linkprio(hdr
, priority
);
1898 msg_set_redundant_link(hdr
, node_up
);
1899 msg_set_seq_gap(hdr
, 0);
1900 msg_set_seqno(hdr
, l
->snd_nxt
+ U16_MAX
/ 2);
1902 if (mtyp
== STATE_MSG
) {
1903 if (l
->peer_caps
& TIPC_LINK_PROTO_SEQNO
)
1904 msg_set_seqno(hdr
, l
->snd_nxt_state
++);
1905 msg_set_seq_gap(hdr
, rcvgap
);
1906 bc_rcvgap
= link_bc_rcv_gap(bcl
);
1907 msg_set_bc_gap(hdr
, bc_rcvgap
);
1908 msg_set_probe(hdr
, probe
);
1909 msg_set_is_keepalive(hdr
, probe
|| probe_reply
);
1910 if (l
->peer_caps
& TIPC_GAP_ACK_BLOCK
)
1911 glen
= tipc_build_gap_ack_blks(l
, hdr
);
1912 tipc_mon_prep(l
->net
, data
+ glen
, &dlen
, mstate
, l
->bearer_id
);
1913 msg_set_size(hdr
, INT_H_SIZE
+ glen
+ dlen
);
1914 skb_trim(skb
, INT_H_SIZE
+ glen
+ dlen
);
1915 l
->stats
.sent_states
++;
1918 /* RESET_MSG or ACTIVATE_MSG */
1919 if (mtyp
== ACTIVATE_MSG
) {
1920 msg_set_dest_session_valid(hdr
, 1);
1921 msg_set_dest_session(hdr
, l
->peer_session
);
1923 msg_set_max_pkt(hdr
, l
->advertised_mtu
);
1924 strcpy(data
, l
->if_name
);
1925 msg_set_size(hdr
, INT_H_SIZE
+ TIPC_MAX_IF_NAME
);
1926 skb_trim(skb
, INT_H_SIZE
+ TIPC_MAX_IF_NAME
);
1929 l
->stats
.sent_probes
++;
1931 l
->stats
.sent_nacks
++;
1933 bcl
->stats
.sent_nacks
++;
1934 skb
->priority
= TC_PRIO_CONTROL
;
1935 __skb_queue_tail(xmitq
, skb
);
1936 trace_tipc_proto_build(skb
, false, l
->name
);
1939 void tipc_link_create_dummy_tnl_msg(struct tipc_link
*l
,
1940 struct sk_buff_head
*xmitq
)
1942 u32 onode
= tipc_own_addr(l
->net
);
1943 struct tipc_msg
*hdr
, *ihdr
;
1944 struct sk_buff_head tnlq
;
1945 struct sk_buff
*skb
;
1946 u32 dnode
= l
->addr
;
1948 __skb_queue_head_init(&tnlq
);
1949 skb
= tipc_msg_create(TUNNEL_PROTOCOL
, FAILOVER_MSG
,
1950 INT_H_SIZE
, BASIC_H_SIZE
,
1951 dnode
, onode
, 0, 0, 0);
1953 pr_warn("%sunable to create tunnel packet\n", link_co_err
);
1958 msg_set_msgcnt(hdr
, 1);
1959 msg_set_bearer_id(hdr
, l
->peer_bearer_id
);
1961 ihdr
= (struct tipc_msg
*)msg_data(hdr
);
1962 tipc_msg_init(onode
, ihdr
, TIPC_LOW_IMPORTANCE
, TIPC_DIRECT_MSG
,
1963 BASIC_H_SIZE
, dnode
);
1964 msg_set_errcode(ihdr
, TIPC_ERR_NO_PORT
);
1965 __skb_queue_tail(&tnlq
, skb
);
1966 tipc_link_xmit(l
, &tnlq
, xmitq
);
1969 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1970 * with contents of the link's transmit and backlog queues.
1972 void tipc_link_tnl_prepare(struct tipc_link
*l
, struct tipc_link
*tnl
,
1973 int mtyp
, struct sk_buff_head
*xmitq
)
1975 struct sk_buff_head
*fdefq
= &tnl
->failover_deferdq
;
1976 struct sk_buff
*skb
, *tnlskb
;
1977 struct tipc_msg
*hdr
, tnlhdr
;
1978 struct sk_buff_head
*queue
= &l
->transmq
;
1979 struct sk_buff_head tmpxq
, tnlq
, frags
;
1980 u16 pktlen
, pktcnt
, seqno
= l
->snd_nxt
;
1981 bool pktcnt_need_update
= false;
1988 __skb_queue_head_init(&tnlq
);
1990 * From now on, send only one single ("dummy") SYNCH message
1991 * to peer. The SYNCH message does not contain any data, just
1992 * a header conveying the synch point to the peer.
1994 if (mtyp
== SYNCH_MSG
&& (tnl
->peer_caps
& TIPC_TUNNEL_ENHANCED
)) {
1995 tnlskb
= tipc_msg_create(TUNNEL_PROTOCOL
, SYNCH_MSG
,
1996 INT_H_SIZE
, 0, l
->addr
,
1997 tipc_own_addr(l
->net
),
2000 pr_warn("%sunable to create dummy SYNCH_MSG\n",
2005 hdr
= buf_msg(tnlskb
);
2006 syncpt
= l
->snd_nxt
+ skb_queue_len(&l
->backlogq
) - 1;
2007 msg_set_syncpt(hdr
, syncpt
);
2008 msg_set_bearer_id(hdr
, l
->peer_bearer_id
);
2009 __skb_queue_tail(&tnlq
, tnlskb
);
2010 tipc_link_xmit(tnl
, &tnlq
, xmitq
);
2014 __skb_queue_head_init(&tmpxq
);
2015 __skb_queue_head_init(&frags
);
2016 /* At least one packet required for safe algorithm => add dummy */
2017 skb
= tipc_msg_create(TIPC_LOW_IMPORTANCE
, TIPC_DIRECT_MSG
,
2018 BASIC_H_SIZE
, 0, l
->addr
, tipc_own_addr(l
->net
),
2019 0, 0, TIPC_ERR_NO_PORT
);
2021 pr_warn("%sunable to create tunnel packet\n", link_co_err
);
2024 __skb_queue_tail(&tnlq
, skb
);
2025 tipc_link_xmit(l
, &tnlq
, &tmpxq
);
2026 __skb_queue_purge(&tmpxq
);
2028 /* Initialize reusable tunnel packet header */
2029 tipc_msg_init(tipc_own_addr(l
->net
), &tnlhdr
, TUNNEL_PROTOCOL
,
2030 mtyp
, INT_H_SIZE
, l
->addr
);
2031 if (mtyp
== SYNCH_MSG
)
2032 pktcnt
= l
->snd_nxt
- buf_seqno(skb_peek(&l
->transmq
));
2034 pktcnt
= skb_queue_len(&l
->transmq
);
2035 pktcnt
+= skb_queue_len(&l
->backlogq
);
2036 msg_set_msgcnt(&tnlhdr
, pktcnt
);
2037 msg_set_bearer_id(&tnlhdr
, l
->peer_bearer_id
);
2039 /* Wrap each packet into a tunnel packet */
2040 skb_queue_walk(queue
, skb
) {
2042 if (queue
== &l
->backlogq
)
2043 msg_set_seqno(hdr
, seqno
++);
2044 pktlen
= msg_size(hdr
);
2046 /* Tunnel link MTU is not large enough? This could be
2048 * 1) Link MTU has just changed or set differently;
2049 * 2) Or FAILOVER on the top of a SYNCH message
2051 * The 2nd case should not happen if peer supports
2052 * TIPC_TUNNEL_ENHANCED
2054 if (pktlen
> tnl
->mtu
- INT_H_SIZE
) {
2055 if (mtyp
== FAILOVER_MSG
&&
2056 (tnl
->peer_caps
& TIPC_TUNNEL_ENHANCED
)) {
2057 rc
= tipc_msg_fragment(skb
, &tnlhdr
, tnl
->mtu
,
2060 pr_warn("%sunable to frag msg: rc %d\n",
2064 pktcnt
+= skb_queue_len(&frags
) - 1;
2065 pktcnt_need_update
= true;
2066 skb_queue_splice_tail_init(&frags
, &tnlq
);
2069 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2070 * => Just warn it and return!
2072 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2073 link_co_err
, msg_user(hdr
),
2074 msg_type(hdr
), msg_size(hdr
));
2078 msg_set_size(&tnlhdr
, pktlen
+ INT_H_SIZE
);
2079 tnlskb
= tipc_buf_acquire(pktlen
+ INT_H_SIZE
, GFP_ATOMIC
);
2081 pr_warn("%sunable to send packet\n", link_co_err
);
2084 skb_copy_to_linear_data(tnlskb
, &tnlhdr
, INT_H_SIZE
);
2085 skb_copy_to_linear_data_offset(tnlskb
, INT_H_SIZE
, hdr
, pktlen
);
2086 __skb_queue_tail(&tnlq
, tnlskb
);
2088 if (queue
!= &l
->backlogq
) {
2089 queue
= &l
->backlogq
;
2093 if (pktcnt_need_update
)
2094 skb_queue_walk(&tnlq
, skb
) {
2096 msg_set_msgcnt(hdr
, pktcnt
);
2099 tipc_link_xmit(tnl
, &tnlq
, xmitq
);
2101 if (mtyp
== FAILOVER_MSG
) {
2102 tnl
->drop_point
= l
->rcv_nxt
;
2103 tnl
->failover_reasm_skb
= l
->reasm_buf
;
2104 l
->reasm_buf
= NULL
;
2106 /* Failover the link's deferdq */
2107 if (unlikely(!skb_queue_empty(fdefq
))) {
2108 pr_warn("Link failover deferdq not empty: %d!\n",
2109 skb_queue_len(fdefq
));
2110 __skb_queue_purge(fdefq
);
2112 skb_queue_splice_init(&l
->deferdq
, fdefq
);
2117 * tipc_link_failover_prepare() - prepare tnl for link failover
2119 * This is a special version of the precursor - tipc_link_tnl_prepare(),
2120 * see the tipc_node_link_failover() for details
2124 * @xmitq: queue for messages to be xmited
2126 void tipc_link_failover_prepare(struct tipc_link
*l
, struct tipc_link
*tnl
,
2127 struct sk_buff_head
*xmitq
)
2129 struct sk_buff_head
*fdefq
= &tnl
->failover_deferdq
;
2131 tipc_link_create_dummy_tnl_msg(tnl
, xmitq
);
2133 /* This failover link endpoint was never established before,
2134 * so it has not received anything from peer.
2135 * Otherwise, it must be a normal failover situation or the
2136 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2137 * would have to start over from scratch instead.
2139 tnl
->drop_point
= 1;
2140 tnl
->failover_reasm_skb
= NULL
;
2142 /* Initiate the link's failover deferdq */
2143 if (unlikely(!skb_queue_empty(fdefq
))) {
2144 pr_warn("Link failover deferdq not empty: %d!\n",
2145 skb_queue_len(fdefq
));
2146 __skb_queue_purge(fdefq
);
2150 /* tipc_link_validate_msg(): validate message against current link state
2151 * Returns true if message should be accepted, otherwise false
2153 bool tipc_link_validate_msg(struct tipc_link
*l
, struct tipc_msg
*hdr
)
2155 u16 curr_session
= l
->peer_session
;
2156 u16 session
= msg_session(hdr
);
2157 int mtyp
= msg_type(hdr
);
2159 if (msg_user(hdr
) != LINK_PROTOCOL
)
2166 /* Accept only RESET with new session number */
2167 return more(session
, curr_session
);
2171 /* Accept only ACTIVATE with new or current session number */
2172 return !less(session
, curr_session
);
2174 /* Accept only STATE with current session number */
2177 if (session
!= curr_session
)
2179 /* Extra sanity check */
2180 if (!link_is_up(l
) && msg_ack(hdr
))
2182 if (!(l
->peer_caps
& TIPC_LINK_PROTO_SEQNO
))
2184 /* Accept only STATE with new sequence number */
2185 return !less(msg_seqno(hdr
), l
->rcv_nxt_state
);
2191 /* tipc_link_proto_rcv(): receive link level protocol message :
2192 * Note that network plane id propagates through the network, and may
2193 * change at any time. The node with lowest numerical id determines
2196 static int tipc_link_proto_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
2197 struct sk_buff_head
*xmitq
)
2199 struct tipc_msg
*hdr
= buf_msg(skb
);
2200 struct tipc_gap_ack_blks
*ga
= NULL
;
2201 bool reply
= msg_probe(hdr
), retransmitted
= false;
2202 u32 dlen
= msg_data_sz(hdr
), glen
= 0;
2203 u16 peers_snd_nxt
= msg_next_sent(hdr
);
2204 u16 peers_tol
= msg_link_tolerance(hdr
);
2205 u16 peers_prio
= msg_linkprio(hdr
);
2206 u16 gap
= msg_seq_gap(hdr
);
2207 u16 ack
= msg_ack(hdr
);
2208 u16 rcv_nxt
= l
->rcv_nxt
;
2210 int mtyp
= msg_type(hdr
);
2211 int rc
= 0, released
;
2215 trace_tipc_proto_rcv(skb
, false, l
->name
);
2220 if (tipc_link_is_blocked(l
) || !xmitq
)
2223 if (tipc_own_addr(l
->net
) > msg_prevnode(hdr
))
2224 l
->net_plane
= msg_net_plane(hdr
);
2228 data
= msg_data(hdr
);
2230 if (!tipc_link_validate_msg(l
, hdr
)) {
2231 trace_tipc_skb_dump(skb
, false, "PROTO invalid (1)!");
2232 trace_tipc_link_dump(l
, TIPC_DUMP_NONE
, "PROTO invalid (1)!");
2239 /* Complete own link name with peer's interface name */
2240 if_name
= strrchr(l
->name
, ':') + 1;
2241 if (sizeof(l
->name
) - (if_name
- l
->name
) <= TIPC_MAX_IF_NAME
)
2243 if (msg_data_sz(hdr
) < TIPC_MAX_IF_NAME
)
2245 strncpy(if_name
, data
, TIPC_MAX_IF_NAME
);
2247 /* Update own tolerance if peer indicates a non-zero value */
2248 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
)) {
2249 l
->tolerance
= peers_tol
;
2250 l
->bc_rcvlink
->tolerance
= peers_tol
;
2252 /* Update own priority if peer's priority is higher */
2253 if (in_range(peers_prio
, l
->priority
+ 1, TIPC_MAX_LINK_PRI
))
2254 l
->priority
= peers_prio
;
2256 /* If peer is going down we want full re-establish cycle */
2257 if (msg_peer_stopping(hdr
)) {
2258 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
2262 /* If this endpoint was re-created while peer was ESTABLISHING
2263 * it doesn't know current session number. Force re-synch.
2265 if (mtyp
== ACTIVATE_MSG
&& msg_dest_session_valid(hdr
) &&
2266 l
->session
!= msg_dest_session(hdr
)) {
2267 if (less(l
->session
, msg_dest_session(hdr
)))
2268 l
->session
= msg_dest_session(hdr
) + 1;
2272 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2273 if (mtyp
== RESET_MSG
|| !link_is_up(l
))
2274 rc
= tipc_link_fsm_evt(l
, LINK_PEER_RESET_EVT
);
2276 /* ACTIVATE_MSG takes up link if it was already locally reset */
2277 if (mtyp
== ACTIVATE_MSG
&& l
->state
== LINK_ESTABLISHING
)
2278 rc
= TIPC_LINK_UP_EVT
;
2280 l
->peer_session
= msg_session(hdr
);
2281 l
->in_session
= true;
2282 l
->peer_bearer_id
= msg_bearer_id(hdr
);
2283 if (l
->mtu
> msg_max_pkt(hdr
))
2284 l
->mtu
= msg_max_pkt(hdr
);
2288 /* Validate Gap ACK blocks, drop if invalid */
2289 glen
= tipc_get_gap_ack_blks(&ga
, l
, hdr
, true);
2293 l
->rcv_nxt_state
= msg_seqno(hdr
) + 1;
2295 /* Update own tolerance if peer indicates a non-zero value */
2296 if (in_range(peers_tol
, TIPC_MIN_LINK_TOL
, TIPC_MAX_LINK_TOL
)) {
2297 l
->tolerance
= peers_tol
;
2298 l
->bc_rcvlink
->tolerance
= peers_tol
;
2300 /* Update own prio if peer indicates a different value */
2301 if ((peers_prio
!= l
->priority
) &&
2302 in_range(peers_prio
, 1, TIPC_MAX_LINK_PRI
)) {
2303 l
->priority
= peers_prio
;
2304 rc
= tipc_link_fsm_evt(l
, LINK_FAILURE_EVT
);
2307 l
->silent_intv_cnt
= 0;
2308 l
->stats
.recv_states
++;
2310 l
->stats
.recv_probes
++;
2312 if (!link_is_up(l
)) {
2313 if (l
->state
== LINK_ESTABLISHING
)
2314 rc
= TIPC_LINK_UP_EVT
;
2318 tipc_mon_rcv(l
->net
, data
+ glen
, dlen
- glen
, l
->addr
,
2319 &l
->mon_state
, l
->bearer_id
);
2321 /* Send NACK if peer has sent pkts we haven't received yet */
2322 if ((reply
|| msg_is_keepalive(hdr
)) &&
2323 more(peers_snd_nxt
, rcv_nxt
) &&
2324 !tipc_link_is_synching(l
) &&
2325 skb_queue_empty(&l
->deferdq
))
2326 rcvgap
= peers_snd_nxt
- l
->rcv_nxt
;
2327 if (rcvgap
|| reply
)
2328 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, reply
,
2329 rcvgap
, 0, 0, xmitq
);
2331 released
= tipc_link_advance_transmq(l
, l
, ack
, gap
, ga
, xmitq
,
2332 &retransmitted
, &rc
);
2334 l
->stats
.recv_nacks
++;
2335 if (released
|| retransmitted
)
2336 tipc_link_update_cwin(l
, released
, retransmitted
);
2338 tipc_link_advance_backlog(l
, xmitq
);
2339 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
2340 link_prepare_wakeup(l
);
2347 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2349 static bool tipc_link_build_bc_proto_msg(struct tipc_link
*l
, bool bcast
,
2351 struct sk_buff_head
*xmitq
)
2353 struct sk_buff
*skb
;
2354 struct tipc_msg
*hdr
;
2355 struct sk_buff
*dfrd_skb
= skb_peek(&l
->deferdq
);
2356 u16 ack
= l
->rcv_nxt
- 1;
2357 u16 gap_to
= peers_snd_nxt
- 1;
2359 skb
= tipc_msg_create(BCAST_PROTOCOL
, STATE_MSG
, INT_H_SIZE
,
2360 0, l
->addr
, tipc_own_addr(l
->net
), 0, 0, 0);
2364 msg_set_last_bcast(hdr
, l
->bc_sndlink
->snd_nxt
- 1);
2365 msg_set_bcast_ack(hdr
, ack
);
2366 msg_set_bcgap_after(hdr
, ack
);
2368 gap_to
= buf_seqno(dfrd_skb
) - 1;
2369 msg_set_bcgap_to(hdr
, gap_to
);
2370 msg_set_non_seq(hdr
, bcast
);
2371 __skb_queue_tail(xmitq
, skb
);
2375 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2377 * Give a newly added peer node the sequence number where it should
2378 * start receiving and acking broadcast packets.
2380 static void tipc_link_build_bc_init_msg(struct tipc_link
*l
,
2381 struct sk_buff_head
*xmitq
)
2383 struct sk_buff_head list
;
2385 __skb_queue_head_init(&list
);
2386 if (!tipc_link_build_bc_proto_msg(l
->bc_rcvlink
, false, 0, &list
))
2388 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list
)), true);
2389 tipc_link_xmit(l
, &list
, xmitq
);
2392 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2394 void tipc_link_bc_init_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
)
2396 int mtyp
= msg_type(hdr
);
2397 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
2402 if (msg_user(hdr
) == BCAST_PROTOCOL
) {
2403 l
->rcv_nxt
= peers_snd_nxt
;
2404 l
->state
= LINK_ESTABLISHED
;
2408 if (l
->peer_caps
& TIPC_BCAST_SYNCH
)
2411 if (msg_peer_node_is_up(hdr
))
2414 /* Compatibility: accept older, less safe initial synch data */
2415 if ((mtyp
== RESET_MSG
) || (mtyp
== ACTIVATE_MSG
))
2416 l
->rcv_nxt
= peers_snd_nxt
;
2419 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2421 int tipc_link_bc_sync_rcv(struct tipc_link
*l
, struct tipc_msg
*hdr
,
2422 struct sk_buff_head
*xmitq
)
2424 u16 peers_snd_nxt
= msg_bc_snd_nxt(hdr
);
2430 if (!msg_peer_node_is_up(hdr
))
2433 /* Open when peer acknowledges our bcast init msg (pkt #1) */
2435 l
->bc_peer_is_up
= true;
2437 if (!l
->bc_peer_is_up
)
2440 /* Ignore if peers_snd_nxt goes beyond receive window */
2441 if (more(peers_snd_nxt
, l
->rcv_nxt
+ l
->window
))
2444 l
->snd_nxt
= peers_snd_nxt
;
2445 if (link_bc_rcv_gap(l
))
2446 rc
|= TIPC_LINK_SND_STATE
;
2448 /* Return now if sender supports nack via STATE messages */
2449 if (l
->peer_caps
& TIPC_BCAST_STATE_NACK
)
2452 /* Otherwise, be backwards compatible */
2454 if (!more(peers_snd_nxt
, l
->rcv_nxt
)) {
2455 l
->nack_state
= BC_NACK_SND_CONDITIONAL
;
2459 /* Don't NACK if one was recently sent or peeked */
2460 if (l
->nack_state
== BC_NACK_SND_SUPPRESS
) {
2461 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
2465 /* Conditionally delay NACK sending until next synch rcv */
2466 if (l
->nack_state
== BC_NACK_SND_CONDITIONAL
) {
2467 l
->nack_state
= BC_NACK_SND_UNCONDITIONAL
;
2468 if ((peers_snd_nxt
- l
->rcv_nxt
) < TIPC_MIN_LINK_WIN
)
2472 /* Send NACK now but suppress next one */
2473 tipc_link_build_bc_proto_msg(l
, true, peers_snd_nxt
, xmitq
);
2474 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
2478 int tipc_link_bc_ack_rcv(struct tipc_link
*r
, u16 acked
, u16 gap
,
2479 struct tipc_gap_ack_blks
*ga
,
2480 struct sk_buff_head
*xmitq
,
2481 struct sk_buff_head
*retrq
)
2483 struct tipc_link
*l
= r
->bc_sndlink
;
2484 bool unused
= false;
2487 if (!link_is_up(r
) || !r
->bc_peer_is_up
)
2491 l
->stats
.recv_nacks
++;
2492 r
->stats
.recv_nacks
++;
2495 if (less(acked
, r
->acked
) || (acked
== r
->acked
&& !gap
&& !ga
))
2498 trace_tipc_link_bc_ack(r
, acked
, gap
, &l
->transmq
);
2499 tipc_link_advance_transmq(l
, r
, acked
, gap
, ga
, retrq
, &unused
, &rc
);
2501 tipc_link_advance_backlog(l
, xmitq
);
2502 if (unlikely(!skb_queue_empty(&l
->wakeupq
)))
2503 link_prepare_wakeup(l
);
2508 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2509 * This function is here for backwards compatibility, since
2510 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2512 int tipc_link_bc_nack_rcv(struct tipc_link
*l
, struct sk_buff
*skb
,
2513 struct sk_buff_head
*xmitq
)
2515 struct tipc_msg
*hdr
= buf_msg(skb
);
2516 u32 dnode
= msg_destnode(hdr
);
2517 int mtyp
= msg_type(hdr
);
2518 u16 acked
= msg_bcast_ack(hdr
);
2519 u16 from
= acked
+ 1;
2520 u16 to
= msg_bcgap_to(hdr
);
2521 u16 peers_snd_nxt
= to
+ 1;
2526 if (!tipc_link_is_up(l
) || !l
->bc_peer_is_up
)
2529 if (mtyp
!= STATE_MSG
)
2532 if (dnode
== tipc_own_addr(l
->net
)) {
2533 rc
= tipc_link_bc_ack_rcv(l
, acked
, to
- acked
, NULL
, xmitq
,
2535 l
->stats
.recv_nacks
++;
2539 /* Msg for other node => suppress own NACK at next sync if applicable */
2540 if (more(peers_snd_nxt
, l
->rcv_nxt
) && !less(l
->rcv_nxt
, from
))
2541 l
->nack_state
= BC_NACK_SND_SUPPRESS
;
2546 void tipc_link_set_queue_limits(struct tipc_link
*l
, u32 min_win
, u32 max_win
)
2548 int max_bulk
= TIPC_MAX_PUBL
/ (l
->mtu
/ ITEM_SIZE
);
2550 l
->min_win
= min_win
;
2551 l
->ssthresh
= max_win
;
2552 l
->max_win
= max_win
;
2553 l
->window
= min_win
;
2554 l
->backlog
[TIPC_LOW_IMPORTANCE
].limit
= min_win
* 2;
2555 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].limit
= min_win
* 4;
2556 l
->backlog
[TIPC_HIGH_IMPORTANCE
].limit
= min_win
* 6;
2557 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].limit
= min_win
* 8;
2558 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].limit
= max_bulk
;
2562 * tipc_link_reset_stats - reset link statistics
2563 * @l: pointer to link
2565 void tipc_link_reset_stats(struct tipc_link
*l
)
2567 memset(&l
->stats
, 0, sizeof(l
->stats
));
2570 static void link_print(struct tipc_link
*l
, const char *str
)
2572 struct sk_buff
*hskb
= skb_peek(&l
->transmq
);
2573 u16 head
= hskb
? msg_seqno(buf_msg(hskb
)) : l
->snd_nxt
- 1;
2574 u16 tail
= l
->snd_nxt
- 1;
2576 pr_info("%s Link <%s> state %x\n", str
, l
->name
, l
->state
);
2577 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2578 skb_queue_len(&l
->transmq
), head
, tail
,
2579 skb_queue_len(&l
->backlogq
), l
->snd_nxt
, l
->rcv_nxt
);
2582 /* Parse and validate nested (link) properties valid for media, bearer and link
2584 int tipc_nl_parse_link_prop(struct nlattr
*prop
, struct nlattr
*props
[])
2588 err
= nla_parse_nested_deprecated(props
, TIPC_NLA_PROP_MAX
, prop
,
2589 tipc_nl_prop_policy
, NULL
);
2593 if (props
[TIPC_NLA_PROP_PRIO
]) {
2596 prio
= nla_get_u32(props
[TIPC_NLA_PROP_PRIO
]);
2597 if (prio
> TIPC_MAX_LINK_PRI
)
2601 if (props
[TIPC_NLA_PROP_TOL
]) {
2604 tol
= nla_get_u32(props
[TIPC_NLA_PROP_TOL
]);
2605 if ((tol
< TIPC_MIN_LINK_TOL
) || (tol
> TIPC_MAX_LINK_TOL
))
2609 if (props
[TIPC_NLA_PROP_WIN
]) {
2612 max_win
= nla_get_u32(props
[TIPC_NLA_PROP_WIN
]);
2613 if (max_win
< TIPC_DEF_LINK_WIN
|| max_win
> TIPC_MAX_LINK_WIN
)
2620 static int __tipc_nl_add_stats(struct sk_buff
*skb
, struct tipc_stats
*s
)
2623 struct nlattr
*stats
;
2630 struct nla_map map
[] = {
2631 {TIPC_NLA_STATS_RX_INFO
, 0},
2632 {TIPC_NLA_STATS_RX_FRAGMENTS
, s
->recv_fragments
},
2633 {TIPC_NLA_STATS_RX_FRAGMENTED
, s
->recv_fragmented
},
2634 {TIPC_NLA_STATS_RX_BUNDLES
, s
->recv_bundles
},
2635 {TIPC_NLA_STATS_RX_BUNDLED
, s
->recv_bundled
},
2636 {TIPC_NLA_STATS_TX_INFO
, 0},
2637 {TIPC_NLA_STATS_TX_FRAGMENTS
, s
->sent_fragments
},
2638 {TIPC_NLA_STATS_TX_FRAGMENTED
, s
->sent_fragmented
},
2639 {TIPC_NLA_STATS_TX_BUNDLES
, s
->sent_bundles
},
2640 {TIPC_NLA_STATS_TX_BUNDLED
, s
->sent_bundled
},
2641 {TIPC_NLA_STATS_MSG_PROF_TOT
, (s
->msg_length_counts
) ?
2642 s
->msg_length_counts
: 1},
2643 {TIPC_NLA_STATS_MSG_LEN_CNT
, s
->msg_length_counts
},
2644 {TIPC_NLA_STATS_MSG_LEN_TOT
, s
->msg_lengths_total
},
2645 {TIPC_NLA_STATS_MSG_LEN_P0
, s
->msg_length_profile
[0]},
2646 {TIPC_NLA_STATS_MSG_LEN_P1
, s
->msg_length_profile
[1]},
2647 {TIPC_NLA_STATS_MSG_LEN_P2
, s
->msg_length_profile
[2]},
2648 {TIPC_NLA_STATS_MSG_LEN_P3
, s
->msg_length_profile
[3]},
2649 {TIPC_NLA_STATS_MSG_LEN_P4
, s
->msg_length_profile
[4]},
2650 {TIPC_NLA_STATS_MSG_LEN_P5
, s
->msg_length_profile
[5]},
2651 {TIPC_NLA_STATS_MSG_LEN_P6
, s
->msg_length_profile
[6]},
2652 {TIPC_NLA_STATS_RX_STATES
, s
->recv_states
},
2653 {TIPC_NLA_STATS_RX_PROBES
, s
->recv_probes
},
2654 {TIPC_NLA_STATS_RX_NACKS
, s
->recv_nacks
},
2655 {TIPC_NLA_STATS_RX_DEFERRED
, s
->deferred_recv
},
2656 {TIPC_NLA_STATS_TX_STATES
, s
->sent_states
},
2657 {TIPC_NLA_STATS_TX_PROBES
, s
->sent_probes
},
2658 {TIPC_NLA_STATS_TX_NACKS
, s
->sent_nacks
},
2659 {TIPC_NLA_STATS_TX_ACKS
, s
->sent_acks
},
2660 {TIPC_NLA_STATS_RETRANSMITTED
, s
->retransmitted
},
2661 {TIPC_NLA_STATS_DUPLICATES
, s
->duplicates
},
2662 {TIPC_NLA_STATS_LINK_CONGS
, s
->link_congs
},
2663 {TIPC_NLA_STATS_MAX_QUEUE
, s
->max_queue_sz
},
2664 {TIPC_NLA_STATS_AVG_QUEUE
, s
->queue_sz_counts
?
2665 (s
->accu_queue_sz
/ s
->queue_sz_counts
) : 0}
2668 stats
= nla_nest_start_noflag(skb
, TIPC_NLA_LINK_STATS
);
2672 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
2673 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
2676 nla_nest_end(skb
, stats
);
2680 nla_nest_cancel(skb
, stats
);
2685 /* Caller should hold appropriate locks to protect the link */
2686 int __tipc_nl_add_link(struct net
*net
, struct tipc_nl_msg
*msg
,
2687 struct tipc_link
*link
, int nlflags
)
2689 u32 self
= tipc_own_addr(net
);
2690 struct nlattr
*attrs
;
2691 struct nlattr
*prop
;
2695 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
2696 nlflags
, TIPC_NL_LINK_GET
);
2700 attrs
= nla_nest_start_noflag(msg
->skb
, TIPC_NLA_LINK
);
2704 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, link
->name
))
2706 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_DEST
, tipc_cluster_mask(self
)))
2708 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_MTU
, link
->mtu
))
2710 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, link
->stats
.recv_pkts
))
2712 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, link
->stats
.sent_pkts
))
2715 if (tipc_link_is_up(link
))
2716 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
2719 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_ACTIVE
))
2722 prop
= nla_nest_start_noflag(msg
->skb
, TIPC_NLA_LINK_PROP
);
2725 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
2727 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_TOL
, link
->tolerance
))
2729 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
,
2732 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_PRIO
, link
->priority
))
2734 nla_nest_end(msg
->skb
, prop
);
2736 err
= __tipc_nl_add_stats(msg
->skb
, &link
->stats
);
2740 nla_nest_end(msg
->skb
, attrs
);
2741 genlmsg_end(msg
->skb
, hdr
);
2746 nla_nest_cancel(msg
->skb
, prop
);
2748 nla_nest_cancel(msg
->skb
, attrs
);
2750 genlmsg_cancel(msg
->skb
, hdr
);
2755 static int __tipc_nl_add_bc_link_stat(struct sk_buff
*skb
,
2756 struct tipc_stats
*stats
)
2759 struct nlattr
*nest
;
2766 struct nla_map map
[] = {
2767 {TIPC_NLA_STATS_RX_INFO
, stats
->recv_pkts
},
2768 {TIPC_NLA_STATS_RX_FRAGMENTS
, stats
->recv_fragments
},
2769 {TIPC_NLA_STATS_RX_FRAGMENTED
, stats
->recv_fragmented
},
2770 {TIPC_NLA_STATS_RX_BUNDLES
, stats
->recv_bundles
},
2771 {TIPC_NLA_STATS_RX_BUNDLED
, stats
->recv_bundled
},
2772 {TIPC_NLA_STATS_TX_INFO
, stats
->sent_pkts
},
2773 {TIPC_NLA_STATS_TX_FRAGMENTS
, stats
->sent_fragments
},
2774 {TIPC_NLA_STATS_TX_FRAGMENTED
, stats
->sent_fragmented
},
2775 {TIPC_NLA_STATS_TX_BUNDLES
, stats
->sent_bundles
},
2776 {TIPC_NLA_STATS_TX_BUNDLED
, stats
->sent_bundled
},
2777 {TIPC_NLA_STATS_RX_NACKS
, stats
->recv_nacks
},
2778 {TIPC_NLA_STATS_RX_DEFERRED
, stats
->deferred_recv
},
2779 {TIPC_NLA_STATS_TX_NACKS
, stats
->sent_nacks
},
2780 {TIPC_NLA_STATS_TX_ACKS
, stats
->sent_acks
},
2781 {TIPC_NLA_STATS_RETRANSMITTED
, stats
->retransmitted
},
2782 {TIPC_NLA_STATS_DUPLICATES
, stats
->duplicates
},
2783 {TIPC_NLA_STATS_LINK_CONGS
, stats
->link_congs
},
2784 {TIPC_NLA_STATS_MAX_QUEUE
, stats
->max_queue_sz
},
2785 {TIPC_NLA_STATS_AVG_QUEUE
, stats
->queue_sz_counts
?
2786 (stats
->accu_queue_sz
/ stats
->queue_sz_counts
) : 0}
2789 nest
= nla_nest_start_noflag(skb
, TIPC_NLA_LINK_STATS
);
2793 for (i
= 0; i
< ARRAY_SIZE(map
); i
++)
2794 if (nla_put_u32(skb
, map
[i
].key
, map
[i
].val
))
2797 nla_nest_end(skb
, nest
);
2801 nla_nest_cancel(skb
, nest
);
2806 int tipc_nl_add_bc_link(struct net
*net
, struct tipc_nl_msg
*msg
,
2807 struct tipc_link
*bcl
)
2811 struct nlattr
*attrs
;
2812 struct nlattr
*prop
;
2813 u32 bc_mode
= tipc_bcast_get_mode(net
);
2814 u32 bc_ratio
= tipc_bcast_get_broadcast_ratio(net
);
2819 tipc_bcast_lock(net
);
2821 hdr
= genlmsg_put(msg
->skb
, msg
->portid
, msg
->seq
, &tipc_genl_family
,
2822 NLM_F_MULTI
, TIPC_NL_LINK_GET
);
2824 tipc_bcast_unlock(net
);
2828 attrs
= nla_nest_start_noflag(msg
->skb
, TIPC_NLA_LINK
);
2832 /* The broadcast link is always up */
2833 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_UP
))
2836 if (nla_put_flag(msg
->skb
, TIPC_NLA_LINK_BROADCAST
))
2838 if (nla_put_string(msg
->skb
, TIPC_NLA_LINK_NAME
, bcl
->name
))
2840 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_RX
, 0))
2842 if (nla_put_u32(msg
->skb
, TIPC_NLA_LINK_TX
, 0))
2845 prop
= nla_nest_start_noflag(msg
->skb
, TIPC_NLA_LINK_PROP
);
2848 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_WIN
, bcl
->max_win
))
2850 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_BROADCAST
, bc_mode
))
2852 if (bc_mode
& BCLINK_MODE_SEL
)
2853 if (nla_put_u32(msg
->skb
, TIPC_NLA_PROP_BROADCAST_RATIO
,
2856 nla_nest_end(msg
->skb
, prop
);
2858 err
= __tipc_nl_add_bc_link_stat(msg
->skb
, &bcl
->stats
);
2862 tipc_bcast_unlock(net
);
2863 nla_nest_end(msg
->skb
, attrs
);
2864 genlmsg_end(msg
->skb
, hdr
);
2869 nla_nest_cancel(msg
->skb
, prop
);
2871 nla_nest_cancel(msg
->skb
, attrs
);
2873 tipc_bcast_unlock(net
);
2874 genlmsg_cancel(msg
->skb
, hdr
);
2879 void tipc_link_set_tolerance(struct tipc_link
*l
, u32 tol
,
2880 struct sk_buff_head
*xmitq
)
2884 l
->bc_rcvlink
->tolerance
= tol
;
2886 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, tol
, 0, xmitq
);
2889 void tipc_link_set_prio(struct tipc_link
*l
, u32 prio
,
2890 struct sk_buff_head
*xmitq
)
2893 tipc_link_build_proto_msg(l
, STATE_MSG
, 0, 0, 0, 0, prio
, xmitq
);
2896 void tipc_link_set_abort_limit(struct tipc_link
*l
, u32 limit
)
2898 l
->abort_limit
= limit
;
2902 * tipc_link_dump - dump TIPC link data
2903 * @l: tipc link to be dumped
2904 * @dqueues: bitmask to decide if any link queue to be dumped?
2905 * - TIPC_DUMP_NONE: don't dump link queues
2906 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2907 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2908 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2909 * - TIPC_DUMP_INPUTQ: dump link input queue
2910 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2911 * - TIPC_DUMP_ALL: dump all the link queues above
2912 * @buf: returned buffer of dump data in format
2914 int tipc_link_dump(struct tipc_link
*l
, u16 dqueues
, char *buf
)
2917 size_t sz
= (dqueues
) ? LINK_LMAX
: LINK_LMIN
;
2918 struct sk_buff_head
*list
;
2919 struct sk_buff
*hskb
, *tskb
;
2923 i
+= scnprintf(buf
, sz
, "link data: (null)\n");
2927 i
+= scnprintf(buf
, sz
, "link data: %x", l
->addr
);
2928 i
+= scnprintf(buf
+ i
, sz
- i
, " %x", l
->state
);
2929 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->in_session
);
2930 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->session
);
2931 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->peer_session
);
2932 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->snd_nxt
);
2933 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->rcv_nxt
);
2934 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->snd_nxt_state
);
2935 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->rcv_nxt_state
);
2936 i
+= scnprintf(buf
+ i
, sz
- i
, " %x", l
->peer_caps
);
2937 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->silent_intv_cnt
);
2938 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->rst_cnt
);
2939 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", 0);
2940 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", 0);
2941 i
+= scnprintf(buf
+ i
, sz
- i
, " %u", l
->acked
);
2944 len
= skb_queue_len(list
);
2945 hskb
= skb_peek(list
);
2946 tskb
= skb_peek_tail(list
);
2947 i
+= scnprintf(buf
+ i
, sz
- i
, " | %u %u %u", len
,
2948 (hskb
) ? msg_seqno(buf_msg(hskb
)) : 0,
2949 (tskb
) ? msg_seqno(buf_msg(tskb
)) : 0);
2952 len
= skb_queue_len(list
);
2953 hskb
= skb_peek(list
);
2954 tskb
= skb_peek_tail(list
);
2955 i
+= scnprintf(buf
+ i
, sz
- i
, " | %u %u %u", len
,
2956 (hskb
) ? msg_seqno(buf_msg(hskb
)) : 0,
2957 (tskb
) ? msg_seqno(buf_msg(tskb
)) : 0);
2959 list
= &l
->backlogq
;
2960 len
= skb_queue_len(list
);
2961 hskb
= skb_peek(list
);
2962 tskb
= skb_peek_tail(list
);
2963 i
+= scnprintf(buf
+ i
, sz
- i
, " | %u %u %u", len
,
2964 (hskb
) ? msg_seqno(buf_msg(hskb
)) : 0,
2965 (tskb
) ? msg_seqno(buf_msg(tskb
)) : 0);
2968 len
= skb_queue_len(list
);
2969 hskb
= skb_peek(list
);
2970 tskb
= skb_peek_tail(list
);
2971 i
+= scnprintf(buf
+ i
, sz
- i
, " | %u %u %u\n", len
,
2972 (hskb
) ? msg_seqno(buf_msg(hskb
)) : 0,
2973 (tskb
) ? msg_seqno(buf_msg(tskb
)) : 0);
2975 if (dqueues
& TIPC_DUMP_TRANSMQ
) {
2976 i
+= scnprintf(buf
+ i
, sz
- i
, "transmq: ");
2977 i
+= tipc_list_dump(&l
->transmq
, false, buf
+ i
);
2979 if (dqueues
& TIPC_DUMP_BACKLOGQ
) {
2980 i
+= scnprintf(buf
+ i
, sz
- i
,
2981 "backlogq: <%u %u %u %u %u>, ",
2982 l
->backlog
[TIPC_LOW_IMPORTANCE
].len
,
2983 l
->backlog
[TIPC_MEDIUM_IMPORTANCE
].len
,
2984 l
->backlog
[TIPC_HIGH_IMPORTANCE
].len
,
2985 l
->backlog
[TIPC_CRITICAL_IMPORTANCE
].len
,
2986 l
->backlog
[TIPC_SYSTEM_IMPORTANCE
].len
);
2987 i
+= tipc_list_dump(&l
->backlogq
, false, buf
+ i
);
2989 if (dqueues
& TIPC_DUMP_DEFERDQ
) {
2990 i
+= scnprintf(buf
+ i
, sz
- i
, "deferdq: ");
2991 i
+= tipc_list_dump(&l
->deferdq
, false, buf
+ i
);
2993 if (dqueues
& TIPC_DUMP_INPUTQ
) {
2994 i
+= scnprintf(buf
+ i
, sz
- i
, "inputq: ");
2995 i
+= tipc_list_dump(l
->inputq
, false, buf
+ i
);
2997 if (dqueues
& TIPC_DUMP_WAKEUP
) {
2998 i
+= scnprintf(buf
+ i
, sz
- i
, "wakeup: ");
2999 i
+= tipc_list_dump(&l
->wakeupq
, false, buf
+ i
);