Merge tag 'trace-v5.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
[linux/fpc-iii.git] / net / tipc / link.c
blob6ae2140eb4f7498f262bcb722f5d365188b20794
1 /*
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
6 * All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
37 #include "core.h"
38 #include "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 #include "trace.h"
47 #include "crypto.h"
49 #include <linux/pkt_sched.h>
51 struct tipc_stats {
52 u32 sent_pkts;
53 u32 recv_pkts;
54 u32 sent_states;
55 u32 recv_states;
56 u32 sent_probes;
57 u32 recv_probes;
58 u32 sent_nacks;
59 u32 recv_nacks;
60 u32 sent_acks;
61 u32 sent_bundled;
62 u32 sent_bundles;
63 u32 recv_bundled;
64 u32 recv_bundles;
65 u32 retransmitted;
66 u32 sent_fragmented;
67 u32 sent_fragments;
68 u32 recv_fragmented;
69 u32 recv_fragments;
70 u32 link_congs; /* # port sends blocked by congestion */
71 u32 deferred_recv;
72 u32 duplicates;
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 */
81 /**
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
86 * @timer: link timer
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
152 struct tipc_link {
153 u32 addr;
154 char name[TIPC_MAX_LINK_NAME];
155 struct net *net;
157 /* Management and link supervision data */
158 u16 peer_session;
159 u16 session;
160 u16 snd_nxt_state;
161 u16 rcv_nxt_state;
162 u32 peer_bearer_id;
163 u32 bearer_id;
164 u32 tolerance;
165 u32 abort_limit;
166 u32 state;
167 u16 peer_caps;
168 bool in_session;
169 bool active;
170 u32 silent_intv_cnt;
171 char if_name[TIPC_MAX_IF_NAME];
172 u32 priority;
173 char net_plane;
174 struct tipc_mon_state mon_state;
175 u16 rst_cnt;
177 /* Failover/synch */
178 u16 drop_point;
179 struct sk_buff *failover_reasm_skb;
180 struct sk_buff_head failover_deferdq;
182 /* Max packet negotiation */
183 u16 mtu;
184 u16 advertised_mtu;
186 /* Sending */
187 struct sk_buff_head transmq;
188 struct sk_buff_head backlogq;
189 struct {
190 u16 len;
191 u16 limit;
192 struct sk_buff *target_bskb;
193 } backlog[5];
194 u16 snd_nxt;
196 /* Reception */
197 u16 rcv_nxt;
198 u32 rcv_unacked;
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;
205 u16 window;
206 u16 min_win;
207 u16 ssthresh;
208 u16 max_win;
209 u16 cong_acks;
210 u16 checkpoint;
212 /* Fragmentation/reassembly */
213 struct sk_buff *reasm_buf;
214 struct sk_buff *reasm_tnlmsg;
216 /* Broadcast */
217 u16 ackers;
218 u16 acked;
219 u16 last_gap;
220 struct tipc_gap_ack_blks *last_ga;
221 struct tipc_link *bc_rcvlink;
222 struct tipc_link *bc_sndlink;
223 u8 nack_state;
224 bool bc_peer_is_up;
226 /* Statistics */
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
238 enum {
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))
247 /* Link FSM states:
249 enum {
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,
281 u16 acked, u16 gap,
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,
286 bool retransmitted);
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)
337 l->active = 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)
347 return l->min_win;
350 int tipc_link_max_win(struct tipc_link *l)
352 return l->max_win;
355 int tipc_link_prio(struct tipc_link *l)
357 return l->priority;
360 unsigned long tipc_link_tolerance(struct tipc_link *l)
362 return l->tolerance;
365 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
367 return l->inputq;
370 char tipc_link_plane(struct tipc_link *l)
372 return l->net_plane;
375 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
377 l->peer_caps = capabilities;
380 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
381 struct tipc_link *uc_l,
382 struct sk_buff_head *xmitq)
384 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
386 snd_l->ackers++;
387 rcv_l->acked = snd_l->snd_nxt - 1;
388 snd_l->state = LINK_ESTABLISHED;
389 tipc_link_build_bc_init_msg(uc_l, xmitq);
392 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
393 struct tipc_link *rcv_l,
394 struct sk_buff_head *xmitq)
396 u16 ack = snd_l->snd_nxt - 1;
398 snd_l->ackers--;
399 rcv_l->bc_peer_is_up = true;
400 rcv_l->state = LINK_ESTABLISHED;
401 tipc_link_bc_ack_rcv(rcv_l, ack, 0, NULL, xmitq, NULL);
402 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
403 tipc_link_reset(rcv_l);
404 rcv_l->state = LINK_RESET;
405 if (!snd_l->ackers) {
406 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
407 tipc_link_reset(snd_l);
408 snd_l->state = LINK_RESET;
409 __skb_queue_purge(xmitq);
413 int tipc_link_bc_peers(struct tipc_link *l)
415 return l->ackers;
418 static u16 link_bc_rcv_gap(struct tipc_link *l)
420 struct sk_buff *skb = skb_peek(&l->deferdq);
421 u16 gap = 0;
423 if (more(l->snd_nxt, l->rcv_nxt))
424 gap = l->snd_nxt - l->rcv_nxt;
425 if (skb)
426 gap = buf_seqno(skb) - l->rcv_nxt;
427 return gap;
430 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
432 l->mtu = mtu;
435 int tipc_link_mtu(struct tipc_link *l)
437 return l->mtu;
440 int tipc_link_mss(struct tipc_link *l)
442 #ifdef CONFIG_TIPC_CRYPTO
443 return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
444 #else
445 return l->mtu - INT_H_SIZE;
446 #endif
449 u16 tipc_link_rcv_nxt(struct tipc_link *l)
451 return l->rcv_nxt;
454 u16 tipc_link_acked(struct tipc_link *l)
456 return l->acked;
459 char *tipc_link_name(struct tipc_link *l)
461 return l->name;
464 u32 tipc_link_state(struct tipc_link *l)
466 return l->state;
470 * tipc_link_create - create a new link
471 * @net: pointer to associated network namespace
472 * @if_name: associated interface name
473 * @bearer_id: id (index) of associated bearer
474 * @tolerance: link tolerance to be used by link
475 * @net_plane: network plane (A,B,c..) this link belongs to
476 * @mtu: mtu to be advertised by link
477 * @priority: priority to be used by link
478 * @min_win: minimal send window to be used by link
479 * @max_win: maximal send window to be used by link
480 * @session: session to be used by link
481 * @peer: node id of peer node
482 * @peer_caps: bitmap describing peer node capabilities
483 * @bc_sndlink: the namespace global link used for broadcast sending
484 * @bc_rcvlink: the peer specific link used for broadcast reception
485 * @inputq: queue to put messages ready for delivery
486 * @namedq: queue to put binding table update messages ready for delivery
487 * @link: return value, pointer to put the created link
488 * @self: local unicast link id
489 * @peer_id: 128-bit ID of peer
491 * Return: true if link was created, otherwise false
493 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
494 int tolerance, char net_plane, u32 mtu, int priority,
495 u32 min_win, u32 max_win, u32 session, u32 self,
496 u32 peer, u8 *peer_id, u16 peer_caps,
497 struct tipc_link *bc_sndlink,
498 struct tipc_link *bc_rcvlink,
499 struct sk_buff_head *inputq,
500 struct sk_buff_head *namedq,
501 struct tipc_link **link)
503 char peer_str[NODE_ID_STR_LEN] = {0,};
504 char self_str[NODE_ID_STR_LEN] = {0,};
505 struct tipc_link *l;
507 l = kzalloc(sizeof(*l), GFP_ATOMIC);
508 if (!l)
509 return false;
510 *link = l;
511 l->session = session;
513 /* Set link name for unicast links only */
514 if (peer_id) {
515 tipc_nodeid2string(self_str, tipc_own_id(net));
516 if (strlen(self_str) > 16)
517 sprintf(self_str, "%x", self);
518 tipc_nodeid2string(peer_str, peer_id);
519 if (strlen(peer_str) > 16)
520 sprintf(peer_str, "%x", peer);
522 /* Peer i/f name will be completed by reset/activate message */
523 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
524 self_str, if_name, peer_str);
526 strcpy(l->if_name, if_name);
527 l->addr = peer;
528 l->peer_caps = peer_caps;
529 l->net = net;
530 l->in_session = false;
531 l->bearer_id = bearer_id;
532 l->tolerance = tolerance;
533 if (bc_rcvlink)
534 bc_rcvlink->tolerance = tolerance;
535 l->net_plane = net_plane;
536 l->advertised_mtu = mtu;
537 l->mtu = mtu;
538 l->priority = priority;
539 tipc_link_set_queue_limits(l, min_win, max_win);
540 l->ackers = 1;
541 l->bc_sndlink = bc_sndlink;
542 l->bc_rcvlink = bc_rcvlink;
543 l->inputq = inputq;
544 l->namedq = namedq;
545 l->state = LINK_RESETTING;
546 __skb_queue_head_init(&l->transmq);
547 __skb_queue_head_init(&l->backlogq);
548 __skb_queue_head_init(&l->deferdq);
549 __skb_queue_head_init(&l->failover_deferdq);
550 skb_queue_head_init(&l->wakeupq);
551 skb_queue_head_init(l->inputq);
552 return true;
556 * tipc_link_bc_create - create new link to be used for broadcast
557 * @net: pointer to associated network namespace
558 * @mtu: mtu to be used initially if no peers
559 * @min_win: minimal send window to be used by link
560 * @max_win: maximal send window to be used by link
561 * @inputq: queue to put messages ready for delivery
562 * @namedq: queue to put binding table update messages ready for delivery
563 * @link: return value, pointer to put the created link
564 * @ownnode: identity of own node
565 * @peer: node id of peer node
566 * @peer_id: 128-bit ID of peer
567 * @peer_caps: bitmap describing peer node capabilities
568 * @bc_sndlink: the namespace global link used for broadcast sending
570 * Return: true if link was created, otherwise false
572 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id,
573 int mtu, u32 min_win, u32 max_win, u16 peer_caps,
574 struct sk_buff_head *inputq,
575 struct sk_buff_head *namedq,
576 struct tipc_link *bc_sndlink,
577 struct tipc_link **link)
579 struct tipc_link *l;
581 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win,
582 max_win, 0, ownnode, peer, NULL, peer_caps,
583 bc_sndlink, NULL, inputq, namedq, link))
584 return false;
586 l = *link;
587 if (peer_id) {
588 char peer_str[NODE_ID_STR_LEN] = {0,};
590 tipc_nodeid2string(peer_str, peer_id);
591 if (strlen(peer_str) > 16)
592 sprintf(peer_str, "%x", peer);
593 /* Broadcast receiver link name: "broadcast-link:<peer>" */
594 snprintf(l->name, sizeof(l->name), "%s:%s", tipc_bclink_name,
595 peer_str);
596 } else {
597 strcpy(l->name, tipc_bclink_name);
599 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
600 tipc_link_reset(l);
601 l->state = LINK_RESET;
602 l->ackers = 0;
603 l->bc_rcvlink = l;
605 /* Broadcast send link is always up */
606 if (link_is_bc_sndlink(l))
607 l->state = LINK_ESTABLISHED;
609 /* Disable replicast if even a single peer doesn't support it */
610 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
611 tipc_bcast_toggle_rcast(net, false);
613 return true;
617 * tipc_link_fsm_evt - link finite state machine
618 * @l: pointer to link
619 * @evt: state machine event to be processed
621 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
623 int rc = 0;
624 int old_state = l->state;
626 switch (l->state) {
627 case LINK_RESETTING:
628 switch (evt) {
629 case LINK_PEER_RESET_EVT:
630 l->state = LINK_PEER_RESET;
631 break;
632 case LINK_RESET_EVT:
633 l->state = LINK_RESET;
634 break;
635 case LINK_FAILURE_EVT:
636 case LINK_FAILOVER_BEGIN_EVT:
637 case LINK_ESTABLISH_EVT:
638 case LINK_FAILOVER_END_EVT:
639 case LINK_SYNCH_BEGIN_EVT:
640 case LINK_SYNCH_END_EVT:
641 default:
642 goto illegal_evt;
644 break;
645 case LINK_RESET:
646 switch (evt) {
647 case LINK_PEER_RESET_EVT:
648 l->state = LINK_ESTABLISHING;
649 break;
650 case LINK_FAILOVER_BEGIN_EVT:
651 l->state = LINK_FAILINGOVER;
652 case LINK_FAILURE_EVT:
653 case LINK_RESET_EVT:
654 case LINK_ESTABLISH_EVT:
655 case LINK_FAILOVER_END_EVT:
656 break;
657 case LINK_SYNCH_BEGIN_EVT:
658 case LINK_SYNCH_END_EVT:
659 default:
660 goto illegal_evt;
662 break;
663 case LINK_PEER_RESET:
664 switch (evt) {
665 case LINK_RESET_EVT:
666 l->state = LINK_ESTABLISHING;
667 break;
668 case LINK_PEER_RESET_EVT:
669 case LINK_ESTABLISH_EVT:
670 case LINK_FAILURE_EVT:
671 break;
672 case LINK_SYNCH_BEGIN_EVT:
673 case LINK_SYNCH_END_EVT:
674 case LINK_FAILOVER_BEGIN_EVT:
675 case LINK_FAILOVER_END_EVT:
676 default:
677 goto illegal_evt;
679 break;
680 case LINK_FAILINGOVER:
681 switch (evt) {
682 case LINK_FAILOVER_END_EVT:
683 l->state = LINK_RESET;
684 break;
685 case LINK_PEER_RESET_EVT:
686 case LINK_RESET_EVT:
687 case LINK_ESTABLISH_EVT:
688 case LINK_FAILURE_EVT:
689 break;
690 case LINK_FAILOVER_BEGIN_EVT:
691 case LINK_SYNCH_BEGIN_EVT:
692 case LINK_SYNCH_END_EVT:
693 default:
694 goto illegal_evt;
696 break;
697 case LINK_ESTABLISHING:
698 switch (evt) {
699 case LINK_ESTABLISH_EVT:
700 l->state = LINK_ESTABLISHED;
701 break;
702 case LINK_FAILOVER_BEGIN_EVT:
703 l->state = LINK_FAILINGOVER;
704 break;
705 case LINK_RESET_EVT:
706 l->state = LINK_RESET;
707 break;
708 case LINK_FAILURE_EVT:
709 case LINK_PEER_RESET_EVT:
710 case LINK_SYNCH_BEGIN_EVT:
711 case LINK_FAILOVER_END_EVT:
712 break;
713 case LINK_SYNCH_END_EVT:
714 default:
715 goto illegal_evt;
717 break;
718 case LINK_ESTABLISHED:
719 switch (evt) {
720 case LINK_PEER_RESET_EVT:
721 l->state = LINK_PEER_RESET;
722 rc |= TIPC_LINK_DOWN_EVT;
723 break;
724 case LINK_FAILURE_EVT:
725 l->state = LINK_RESETTING;
726 rc |= TIPC_LINK_DOWN_EVT;
727 break;
728 case LINK_RESET_EVT:
729 l->state = LINK_RESET;
730 break;
731 case LINK_ESTABLISH_EVT:
732 case LINK_SYNCH_END_EVT:
733 break;
734 case LINK_SYNCH_BEGIN_EVT:
735 l->state = LINK_SYNCHING;
736 break;
737 case LINK_FAILOVER_BEGIN_EVT:
738 case LINK_FAILOVER_END_EVT:
739 default:
740 goto illegal_evt;
742 break;
743 case LINK_SYNCHING:
744 switch (evt) {
745 case LINK_PEER_RESET_EVT:
746 l->state = LINK_PEER_RESET;
747 rc |= TIPC_LINK_DOWN_EVT;
748 break;
749 case LINK_FAILURE_EVT:
750 l->state = LINK_RESETTING;
751 rc |= TIPC_LINK_DOWN_EVT;
752 break;
753 case LINK_RESET_EVT:
754 l->state = LINK_RESET;
755 break;
756 case LINK_ESTABLISH_EVT:
757 case LINK_SYNCH_BEGIN_EVT:
758 break;
759 case LINK_SYNCH_END_EVT:
760 l->state = LINK_ESTABLISHED;
761 break;
762 case LINK_FAILOVER_BEGIN_EVT:
763 case LINK_FAILOVER_END_EVT:
764 default:
765 goto illegal_evt;
767 break;
768 default:
769 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
771 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
772 return rc;
773 illegal_evt:
774 pr_err("Illegal FSM event %x in state %x on link %s\n",
775 evt, l->state, l->name);
776 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
777 return rc;
780 /* link_profile_stats - update statistical profiling of traffic
782 static void link_profile_stats(struct tipc_link *l)
784 struct sk_buff *skb;
785 struct tipc_msg *msg;
786 int length;
788 /* Update counters used in statistical profiling of send traffic */
789 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
790 l->stats.queue_sz_counts++;
792 skb = skb_peek(&l->transmq);
793 if (!skb)
794 return;
795 msg = buf_msg(skb);
796 length = msg_size(msg);
798 if (msg_user(msg) == MSG_FRAGMENTER) {
799 if (msg_type(msg) != FIRST_FRAGMENT)
800 return;
801 length = msg_size(msg_inner_hdr(msg));
803 l->stats.msg_lengths_total += length;
804 l->stats.msg_length_counts++;
805 if (length <= 64)
806 l->stats.msg_length_profile[0]++;
807 else if (length <= 256)
808 l->stats.msg_length_profile[1]++;
809 else if (length <= 1024)
810 l->stats.msg_length_profile[2]++;
811 else if (length <= 4096)
812 l->stats.msg_length_profile[3]++;
813 else if (length <= 16384)
814 l->stats.msg_length_profile[4]++;
815 else if (length <= 32768)
816 l->stats.msg_length_profile[5]++;
817 else
818 l->stats.msg_length_profile[6]++;
822 * tipc_link_too_silent - check if link is "too silent"
823 * @l: tipc link to be checked
825 * Return: true if the link 'silent_intv_cnt' is about to reach the
826 * 'abort_limit' value, otherwise false
828 bool tipc_link_too_silent(struct tipc_link *l)
830 return (l->silent_intv_cnt + 2 > l->abort_limit);
833 /* tipc_link_timeout - perform periodic task as instructed from node timeout
835 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
837 int mtyp = 0;
838 int rc = 0;
839 bool state = false;
840 bool probe = false;
841 bool setup = false;
842 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
843 u16 bc_acked = l->bc_rcvlink->acked;
844 struct tipc_mon_state *mstate = &l->mon_state;
846 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
847 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
848 switch (l->state) {
849 case LINK_ESTABLISHED:
850 case LINK_SYNCHING:
851 mtyp = STATE_MSG;
852 link_profile_stats(l);
853 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
854 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
855 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
856 state = bc_acked != bc_snt;
857 state |= l->bc_rcvlink->rcv_unacked;
858 state |= l->rcv_unacked;
859 state |= !skb_queue_empty(&l->transmq);
860 probe = mstate->probing;
861 probe |= l->silent_intv_cnt;
862 if (probe || mstate->monitoring)
863 l->silent_intv_cnt++;
864 probe |= !skb_queue_empty(&l->deferdq);
865 if (l->snd_nxt == l->checkpoint) {
866 tipc_link_update_cwin(l, 0, 0);
867 probe = true;
869 l->checkpoint = l->snd_nxt;
870 break;
871 case LINK_RESET:
872 setup = l->rst_cnt++ <= 4;
873 setup |= !(l->rst_cnt % 16);
874 mtyp = RESET_MSG;
875 break;
876 case LINK_ESTABLISHING:
877 setup = true;
878 mtyp = ACTIVATE_MSG;
879 break;
880 case LINK_PEER_RESET:
881 case LINK_RESETTING:
882 case LINK_FAILINGOVER:
883 break;
884 default:
885 break;
888 if (state || probe || setup)
889 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
891 return rc;
895 * link_schedule_user - schedule a message sender for wakeup after congestion
896 * @l: congested link
897 * @hdr: header of message that is being sent
898 * Create pseudo msg to send back to user when congestion abates
900 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
902 u32 dnode = tipc_own_addr(l->net);
903 u32 dport = msg_origport(hdr);
904 struct sk_buff *skb;
906 /* Create and schedule wakeup pseudo message */
907 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
908 dnode, l->addr, dport, 0, 0);
909 if (!skb)
910 return -ENOBUFS;
911 msg_set_dest_droppable(buf_msg(skb), true);
912 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
913 skb_queue_tail(&l->wakeupq, skb);
914 l->stats.link_congs++;
915 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
916 return -ELINKCONG;
920 * link_prepare_wakeup - prepare users for wakeup after congestion
921 * @l: congested link
922 * Wake up a number of waiting users, as permitted by available space
923 * in the send queue
925 static void link_prepare_wakeup(struct tipc_link *l)
927 struct sk_buff_head *wakeupq = &l->wakeupq;
928 struct sk_buff_head *inputq = l->inputq;
929 struct sk_buff *skb, *tmp;
930 struct sk_buff_head tmpq;
931 int avail[5] = {0,};
932 int imp = 0;
934 __skb_queue_head_init(&tmpq);
936 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
937 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
939 skb_queue_walk_safe(wakeupq, skb, tmp) {
940 imp = TIPC_SKB_CB(skb)->chain_imp;
941 if (avail[imp] <= 0)
942 continue;
943 avail[imp]--;
944 __skb_unlink(skb, wakeupq);
945 __skb_queue_tail(&tmpq, skb);
948 spin_lock_bh(&inputq->lock);
949 skb_queue_splice_tail(&tmpq, inputq);
950 spin_unlock_bh(&inputq->lock);
955 * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
956 * the given skb should be next attempted
957 * @skb: skb to set a future retransmission time for
958 * @l: link the skb will be transmitted on
960 static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
961 struct tipc_link *l)
963 if (link_is_bc_sndlink(l))
964 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
965 else
966 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
969 void tipc_link_reset(struct tipc_link *l)
971 struct sk_buff_head list;
972 u32 imp;
974 __skb_queue_head_init(&list);
976 l->in_session = false;
977 /* Force re-synch of peer session number before establishing */
978 l->peer_session--;
979 l->session++;
980 l->mtu = l->advertised_mtu;
982 spin_lock_bh(&l->wakeupq.lock);
983 skb_queue_splice_init(&l->wakeupq, &list);
984 spin_unlock_bh(&l->wakeupq.lock);
986 spin_lock_bh(&l->inputq->lock);
987 skb_queue_splice_init(&list, l->inputq);
988 spin_unlock_bh(&l->inputq->lock);
990 __skb_queue_purge(&l->transmq);
991 __skb_queue_purge(&l->deferdq);
992 __skb_queue_purge(&l->backlogq);
993 __skb_queue_purge(&l->failover_deferdq);
994 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
995 l->backlog[imp].len = 0;
996 l->backlog[imp].target_bskb = NULL;
998 kfree_skb(l->reasm_buf);
999 kfree_skb(l->reasm_tnlmsg);
1000 kfree_skb(l->failover_reasm_skb);
1001 l->reasm_buf = NULL;
1002 l->reasm_tnlmsg = NULL;
1003 l->failover_reasm_skb = NULL;
1004 l->rcv_unacked = 0;
1005 l->snd_nxt = 1;
1006 l->rcv_nxt = 1;
1007 l->snd_nxt_state = 1;
1008 l->rcv_nxt_state = 1;
1009 l->acked = 0;
1010 l->last_gap = 0;
1011 kfree(l->last_ga);
1012 l->last_ga = NULL;
1013 l->silent_intv_cnt = 0;
1014 l->rst_cnt = 0;
1015 l->bc_peer_is_up = false;
1016 memset(&l->mon_state, 0, sizeof(l->mon_state));
1017 tipc_link_reset_stats(l);
1021 * tipc_link_xmit(): enqueue buffer list according to queue situation
1022 * @l: link to use
1023 * @list: chain of buffers containing message
1024 * @xmitq: returned list of packets to be sent by caller
1026 * Consumes the buffer chain.
1027 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
1028 * Return: 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
1030 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
1031 struct sk_buff_head *xmitq)
1033 struct tipc_msg *hdr = buf_msg(skb_peek(list));
1034 struct sk_buff_head *backlogq = &l->backlogq;
1035 struct sk_buff_head *transmq = &l->transmq;
1036 struct sk_buff *skb, *_skb;
1037 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1038 u16 ack = l->rcv_nxt - 1;
1039 u16 seqno = l->snd_nxt;
1040 int pkt_cnt = skb_queue_len(list);
1041 int imp = msg_importance(hdr);
1042 unsigned int mss = tipc_link_mss(l);
1043 unsigned int cwin = l->window;
1044 unsigned int mtu = l->mtu;
1045 bool new_bundle;
1046 int rc = 0;
1048 if (unlikely(msg_size(hdr) > mtu)) {
1049 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1050 skb_queue_len(list), msg_user(hdr),
1051 msg_type(hdr), msg_size(hdr), mtu);
1052 __skb_queue_purge(list);
1053 return -EMSGSIZE;
1056 /* Allow oversubscription of one data msg per source at congestion */
1057 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
1058 if (imp == TIPC_SYSTEM_IMPORTANCE) {
1059 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
1060 return -ENOBUFS;
1062 rc = link_schedule_user(l, hdr);
1065 if (pkt_cnt > 1) {
1066 l->stats.sent_fragmented++;
1067 l->stats.sent_fragments += pkt_cnt;
1070 /* Prepare each packet for sending, and add to relevant queue: */
1071 while ((skb = __skb_dequeue(list))) {
1072 if (likely(skb_queue_len(transmq) < cwin)) {
1073 hdr = buf_msg(skb);
1074 msg_set_seqno(hdr, seqno);
1075 msg_set_ack(hdr, ack);
1076 msg_set_bcast_ack(hdr, bc_ack);
1077 _skb = skb_clone(skb, GFP_ATOMIC);
1078 if (!_skb) {
1079 kfree_skb(skb);
1080 __skb_queue_purge(list);
1081 return -ENOBUFS;
1083 __skb_queue_tail(transmq, skb);
1084 tipc_link_set_skb_retransmit_time(skb, l);
1085 __skb_queue_tail(xmitq, _skb);
1086 TIPC_SKB_CB(skb)->ackers = l->ackers;
1087 l->rcv_unacked = 0;
1088 l->stats.sent_pkts++;
1089 seqno++;
1090 continue;
1092 if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1093 mss, l->addr, &new_bundle)) {
1094 if (skb) {
1095 /* Keep a ref. to the skb for next try */
1096 l->backlog[imp].target_bskb = skb;
1097 l->backlog[imp].len++;
1098 __skb_queue_tail(backlogq, skb);
1099 } else {
1100 if (new_bundle) {
1101 l->stats.sent_bundles++;
1102 l->stats.sent_bundled++;
1104 l->stats.sent_bundled++;
1106 continue;
1108 l->backlog[imp].target_bskb = NULL;
1109 l->backlog[imp].len += (1 + skb_queue_len(list));
1110 __skb_queue_tail(backlogq, skb);
1111 skb_queue_splice_tail_init(list, backlogq);
1113 l->snd_nxt = seqno;
1114 return rc;
1117 static void tipc_link_update_cwin(struct tipc_link *l, int released,
1118 bool retransmitted)
1120 int bklog_len = skb_queue_len(&l->backlogq);
1121 struct sk_buff_head *txq = &l->transmq;
1122 int txq_len = skb_queue_len(txq);
1123 u16 cwin = l->window;
1125 /* Enter fast recovery */
1126 if (unlikely(retransmitted)) {
1127 l->ssthresh = max_t(u16, l->window / 2, 300);
1128 l->window = min_t(u16, l->ssthresh, l->window);
1129 return;
1131 /* Enter slow start */
1132 if (unlikely(!released)) {
1133 l->ssthresh = max_t(u16, l->window / 2, 300);
1134 l->window = l->min_win;
1135 return;
1137 /* Don't increase window if no pressure on the transmit queue */
1138 if (txq_len + bklog_len < cwin)
1139 return;
1141 /* Don't increase window if there are holes the transmit queue */
1142 if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len)
1143 return;
1145 l->cong_acks += released;
1147 /* Slow start */
1148 if (cwin <= l->ssthresh) {
1149 l->window = min_t(u16, cwin + released, l->max_win);
1150 return;
1152 /* Congestion avoidance */
1153 if (l->cong_acks < cwin)
1154 return;
1155 l->window = min_t(u16, ++cwin, l->max_win);
1156 l->cong_acks = 0;
1159 static void tipc_link_advance_backlog(struct tipc_link *l,
1160 struct sk_buff_head *xmitq)
1162 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1163 struct sk_buff_head *txq = &l->transmq;
1164 struct sk_buff *skb, *_skb;
1165 u16 ack = l->rcv_nxt - 1;
1166 u16 seqno = l->snd_nxt;
1167 struct tipc_msg *hdr;
1168 u16 cwin = l->window;
1169 u32 imp;
1171 while (skb_queue_len(txq) < cwin) {
1172 skb = skb_peek(&l->backlogq);
1173 if (!skb)
1174 break;
1175 _skb = skb_clone(skb, GFP_ATOMIC);
1176 if (!_skb)
1177 break;
1178 __skb_dequeue(&l->backlogq);
1179 hdr = buf_msg(skb);
1180 imp = msg_importance(hdr);
1181 l->backlog[imp].len--;
1182 if (unlikely(skb == l->backlog[imp].target_bskb))
1183 l->backlog[imp].target_bskb = NULL;
1184 __skb_queue_tail(&l->transmq, skb);
1185 tipc_link_set_skb_retransmit_time(skb, l);
1187 __skb_queue_tail(xmitq, _skb);
1188 TIPC_SKB_CB(skb)->ackers = l->ackers;
1189 msg_set_seqno(hdr, seqno);
1190 msg_set_ack(hdr, ack);
1191 msg_set_bcast_ack(hdr, bc_ack);
1192 l->rcv_unacked = 0;
1193 l->stats.sent_pkts++;
1194 seqno++;
1196 l->snd_nxt = seqno;
1200 * link_retransmit_failure() - Detect repeated retransmit failures
1201 * @l: tipc link sender
1202 * @r: tipc link receiver (= l in case of unicast)
1203 * @rc: returned code
1205 * Return: true if the repeated retransmit failures happens, otherwise
1206 * false
1208 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1209 int *rc)
1211 struct sk_buff *skb = skb_peek(&l->transmq);
1212 struct tipc_msg *hdr;
1214 if (!skb)
1215 return false;
1217 if (!TIPC_SKB_CB(skb)->retr_cnt)
1218 return false;
1220 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1221 msecs_to_jiffies(r->tolerance * 10)))
1222 return false;
1224 hdr = buf_msg(skb);
1225 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1226 return false;
1228 pr_warn("Retransmission failure on link <%s>\n", l->name);
1229 link_print(l, "State of link ");
1230 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1231 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1232 pr_info("sqno %u, prev: %x, dest: %x\n",
1233 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1234 pr_info("retr_stamp %d, retr_cnt %d\n",
1235 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1236 TIPC_SKB_CB(skb)->retr_cnt);
1238 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1239 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1240 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1242 if (link_is_bc_sndlink(l)) {
1243 r->state = LINK_RESET;
1244 *rc |= TIPC_LINK_DOWN_EVT;
1245 } else {
1246 *rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1249 return true;
1252 /* tipc_data_input - deliver data and name distr msgs to upper layer
1254 * Consumes buffer if message is of right type
1255 * Node lock must be held
1257 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1258 struct sk_buff_head *inputq)
1260 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1261 struct tipc_msg *hdr = buf_msg(skb);
1263 switch (msg_user(hdr)) {
1264 case TIPC_LOW_IMPORTANCE:
1265 case TIPC_MEDIUM_IMPORTANCE:
1266 case TIPC_HIGH_IMPORTANCE:
1267 case TIPC_CRITICAL_IMPORTANCE:
1268 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1269 skb_queue_tail(mc_inputq, skb);
1270 return true;
1272 fallthrough;
1273 case CONN_MANAGER:
1274 skb_queue_tail(inputq, skb);
1275 return true;
1276 case GROUP_PROTOCOL:
1277 skb_queue_tail(mc_inputq, skb);
1278 return true;
1279 case NAME_DISTRIBUTOR:
1280 l->bc_rcvlink->state = LINK_ESTABLISHED;
1281 skb_queue_tail(l->namedq, skb);
1282 return true;
1283 case MSG_BUNDLER:
1284 case TUNNEL_PROTOCOL:
1285 case MSG_FRAGMENTER:
1286 case BCAST_PROTOCOL:
1287 return false;
1288 #ifdef CONFIG_TIPC_CRYPTO
1289 case MSG_CRYPTO:
1290 tipc_crypto_msg_rcv(l->net, skb);
1291 return true;
1292 #endif
1293 default:
1294 pr_warn("Dropping received illegal msg type\n");
1295 kfree_skb(skb);
1296 return true;
1300 /* tipc_link_input - process packet that has passed link protocol check
1302 * Consumes buffer
1304 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1305 struct sk_buff_head *inputq,
1306 struct sk_buff **reasm_skb)
1308 struct tipc_msg *hdr = buf_msg(skb);
1309 struct sk_buff *iskb;
1310 struct sk_buff_head tmpq;
1311 int usr = msg_user(hdr);
1312 int pos = 0;
1314 if (usr == MSG_BUNDLER) {
1315 skb_queue_head_init(&tmpq);
1316 l->stats.recv_bundles++;
1317 l->stats.recv_bundled += msg_msgcnt(hdr);
1318 while (tipc_msg_extract(skb, &iskb, &pos))
1319 tipc_data_input(l, iskb, &tmpq);
1320 tipc_skb_queue_splice_tail(&tmpq, inputq);
1321 return 0;
1322 } else if (usr == MSG_FRAGMENTER) {
1323 l->stats.recv_fragments++;
1324 if (tipc_buf_append(reasm_skb, &skb)) {
1325 l->stats.recv_fragmented++;
1326 tipc_data_input(l, skb, inputq);
1327 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1328 pr_warn_ratelimited("Unable to build fragment list\n");
1329 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1331 return 0;
1332 } else if (usr == BCAST_PROTOCOL) {
1333 tipc_bcast_lock(l->net);
1334 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1335 tipc_bcast_unlock(l->net);
1338 kfree_skb(skb);
1339 return 0;
1342 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1343 * inner message along with the ones in the old link's
1344 * deferdq
1345 * @l: tunnel link
1346 * @skb: TUNNEL_PROTOCOL message
1347 * @inputq: queue to put messages ready for delivery
1349 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1350 struct sk_buff_head *inputq)
1352 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1353 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1354 struct sk_buff_head *fdefq = &l->failover_deferdq;
1355 struct tipc_msg *hdr = buf_msg(skb);
1356 struct sk_buff *iskb;
1357 int ipos = 0;
1358 int rc = 0;
1359 u16 seqno;
1361 if (msg_type(hdr) == SYNCH_MSG) {
1362 kfree_skb(skb);
1363 return 0;
1366 /* Not a fragment? */
1367 if (likely(!msg_nof_fragms(hdr))) {
1368 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1369 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1370 skb_queue_len(fdefq));
1371 return 0;
1373 kfree_skb(skb);
1374 } else {
1375 /* Set fragment type for buf_append */
1376 if (msg_fragm_no(hdr) == 1)
1377 msg_set_type(hdr, FIRST_FRAGMENT);
1378 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1379 msg_set_type(hdr, FRAGMENT);
1380 else
1381 msg_set_type(hdr, LAST_FRAGMENT);
1383 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1384 /* Successful but non-complete reassembly? */
1385 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1386 return 0;
1387 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1388 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1390 iskb = skb;
1393 do {
1394 seqno = buf_seqno(iskb);
1395 if (unlikely(less(seqno, l->drop_point))) {
1396 kfree_skb(iskb);
1397 continue;
1399 if (unlikely(seqno != l->drop_point)) {
1400 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1401 continue;
1404 l->drop_point++;
1405 if (!tipc_data_input(l, iskb, inputq))
1406 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1407 if (unlikely(rc))
1408 break;
1409 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1411 return rc;
1415 * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1416 * @ga: returned pointer to the Gap ACK blocks if any
1417 * @l: the tipc link
1418 * @hdr: the PROTOCOL/STATE_MSG header
1419 * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1421 * Return: the total Gap ACK blocks size
1423 u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1424 struct tipc_msg *hdr, bool uc)
1426 struct tipc_gap_ack_blks *p;
1427 u16 sz = 0;
1429 /* Does peer support the Gap ACK blocks feature? */
1430 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1431 p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1432 sz = ntohs(p->len);
1433 /* Sanity check */
1434 if (sz == struct_size(p, gacks, p->ugack_cnt + p->bgack_cnt)) {
1435 /* Good, check if the desired type exists */
1436 if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1437 goto ok;
1438 /* Backward compatible: peer might not support bc, but uc? */
1439 } else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1440 if (p->ugack_cnt) {
1441 p->bgack_cnt = 0;
1442 goto ok;
1446 /* Other cases: ignore! */
1447 p = NULL;
1450 *ga = p;
1451 return sz;
1454 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1455 struct tipc_link *l, u8 start_index)
1457 struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1458 struct sk_buff *skb = skb_peek(&l->deferdq);
1459 u16 expect, seqno = 0;
1460 u8 n = 0;
1462 if (!skb)
1463 return 0;
1465 expect = buf_seqno(skb);
1466 skb_queue_walk(&l->deferdq, skb) {
1467 seqno = buf_seqno(skb);
1468 if (unlikely(more(seqno, expect))) {
1469 gacks[n].ack = htons(expect - 1);
1470 gacks[n].gap = htons(seqno - expect);
1471 if (++n >= MAX_GAP_ACK_BLKS / 2) {
1472 pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1473 l->name, n,
1474 skb_queue_len(&l->deferdq));
1475 return n;
1477 } else if (unlikely(less(seqno, expect))) {
1478 pr_warn("Unexpected skb in deferdq!\n");
1479 continue;
1481 expect = seqno + 1;
1484 /* last block */
1485 gacks[n].ack = htons(seqno);
1486 gacks[n].gap = 0;
1487 n++;
1488 return n;
1491 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1492 * @l: tipc unicast link
1493 * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1495 * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1496 * links of a certain peer, the buffer after built has the network data format
1497 * as found at the struct tipc_gap_ack_blks definition.
1499 * returns the actual allocated memory size
1501 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1503 struct tipc_link *bcl = l->bc_rcvlink;
1504 struct tipc_gap_ack_blks *ga;
1505 u16 len;
1507 ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1509 /* Start with broadcast link first */
1510 tipc_bcast_lock(bcl->net);
1511 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1512 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1513 ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1514 tipc_bcast_unlock(bcl->net);
1516 /* Now for unicast link, but an explicit NACK only (???) */
1517 ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1518 __tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1520 /* Total len */
1521 len = struct_size(ga, gacks, ga->bgack_cnt + ga->ugack_cnt);
1522 ga->len = htons(len);
1523 return len;
1526 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1527 * acked packets, also doing retransmissions if
1528 * gaps found
1529 * @l: tipc link with transmq queue to be advanced
1530 * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1531 * @acked: seqno of last packet acked by peer without any gaps before
1532 * @gap: # of gap packets
1533 * @ga: buffer pointer to Gap ACK blocks from peer
1534 * @xmitq: queue for accumulating the retransmitted packets if any
1535 * @retransmitted: returned boolean value if a retransmission is really issued
1536 * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1537 * happens (- unlikely case)
1539 * Return: the number of packets released from the link transmq
1541 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1542 u16 acked, u16 gap,
1543 struct tipc_gap_ack_blks *ga,
1544 struct sk_buff_head *xmitq,
1545 bool *retransmitted, int *rc)
1547 struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1548 struct tipc_gap_ack *gacks = NULL;
1549 struct sk_buff *skb, *_skb, *tmp;
1550 struct tipc_msg *hdr;
1551 u32 qlen = skb_queue_len(&l->transmq);
1552 u16 nacked = acked, ngap = gap, gack_cnt = 0;
1553 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1554 u16 ack = l->rcv_nxt - 1;
1555 u16 seqno, n = 0;
1556 u16 end = r->acked, start = end, offset = r->last_gap;
1557 u16 si = (last_ga) ? last_ga->start_index : 0;
1558 bool is_uc = !link_is_bc_sndlink(l);
1559 bool bc_has_acked = false;
1561 trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1563 /* Determine Gap ACK blocks if any for the particular link */
1564 if (ga && is_uc) {
1565 /* Get the Gap ACKs, uc part */
1566 gack_cnt = ga->ugack_cnt;
1567 gacks = &ga->gacks[ga->bgack_cnt];
1568 } else if (ga) {
1569 /* Copy the Gap ACKs, bc part, for later renewal if needed */
1570 this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1571 GFP_ATOMIC);
1572 if (likely(this_ga)) {
1573 this_ga->start_index = 0;
1574 /* Start with the bc Gap ACKs */
1575 gack_cnt = this_ga->bgack_cnt;
1576 gacks = &this_ga->gacks[0];
1577 } else {
1578 /* Hmm, we can get in trouble..., simply ignore it */
1579 pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1583 /* Advance the link transmq */
1584 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1585 seqno = buf_seqno(skb);
1587 next_gap_ack:
1588 if (less_eq(seqno, nacked)) {
1589 if (is_uc)
1590 goto release;
1591 /* Skip packets peer has already acked */
1592 if (!more(seqno, r->acked))
1593 continue;
1594 /* Get the next of last Gap ACK blocks */
1595 while (more(seqno, end)) {
1596 if (!last_ga || si >= last_ga->bgack_cnt)
1597 break;
1598 start = end + offset + 1;
1599 end = ntohs(last_ga->gacks[si].ack);
1600 offset = ntohs(last_ga->gacks[si].gap);
1601 si++;
1602 WARN_ONCE(more(start, end) ||
1603 (!offset &&
1604 si < last_ga->bgack_cnt) ||
1605 si > MAX_GAP_ACK_BLKS,
1606 "Corrupted Gap ACK: %d %d %d %d %d\n",
1607 start, end, offset, si,
1608 last_ga->bgack_cnt);
1610 /* Check against the last Gap ACK block */
1611 if (in_range(seqno, start, end))
1612 continue;
1613 /* Update/release the packet peer is acking */
1614 bc_has_acked = true;
1615 if (--TIPC_SKB_CB(skb)->ackers)
1616 continue;
1617 release:
1618 /* release skb */
1619 __skb_unlink(skb, &l->transmq);
1620 kfree_skb(skb);
1621 } else if (less_eq(seqno, nacked + ngap)) {
1622 /* First gap: check if repeated retrans failures? */
1623 if (unlikely(seqno == acked + 1 &&
1624 link_retransmit_failure(l, r, rc))) {
1625 /* Ignore this bc Gap ACKs if any */
1626 kfree(this_ga);
1627 this_ga = NULL;
1628 break;
1630 /* retransmit skb if unrestricted*/
1631 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1632 continue;
1633 tipc_link_set_skb_retransmit_time(skb, l);
1634 _skb = pskb_copy(skb, GFP_ATOMIC);
1635 if (!_skb)
1636 continue;
1637 hdr = buf_msg(_skb);
1638 msg_set_ack(hdr, ack);
1639 msg_set_bcast_ack(hdr, bc_ack);
1640 _skb->priority = TC_PRIO_CONTROL;
1641 __skb_queue_tail(xmitq, _skb);
1642 l->stats.retransmitted++;
1643 if (!is_uc)
1644 r->stats.retransmitted++;
1645 *retransmitted = true;
1646 /* Increase actual retrans counter & mark first time */
1647 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1648 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1649 } else {
1650 /* retry with Gap ACK blocks if any */
1651 if (n >= gack_cnt)
1652 break;
1653 nacked = ntohs(gacks[n].ack);
1654 ngap = ntohs(gacks[n].gap);
1655 n++;
1656 goto next_gap_ack;
1660 /* Renew last Gap ACK blocks for bc if needed */
1661 if (bc_has_acked) {
1662 if (this_ga) {
1663 kfree(last_ga);
1664 r->last_ga = this_ga;
1665 r->last_gap = gap;
1666 } else if (last_ga) {
1667 if (less(acked, start)) {
1668 si--;
1669 offset = start - acked - 1;
1670 } else if (less(acked, end)) {
1671 acked = end;
1673 if (si < last_ga->bgack_cnt) {
1674 last_ga->start_index = si;
1675 r->last_gap = offset;
1676 } else {
1677 kfree(last_ga);
1678 r->last_ga = NULL;
1679 r->last_gap = 0;
1681 } else {
1682 r->last_gap = 0;
1684 r->acked = acked;
1685 } else {
1686 kfree(this_ga);
1689 return qlen - skb_queue_len(&l->transmq);
1692 /* tipc_link_build_state_msg: prepare link state message for transmission
1694 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1695 * risk of ack storms towards the sender
1697 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1699 if (!l)
1700 return 0;
1702 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1703 if (link_is_bc_rcvlink(l)) {
1704 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1705 return 0;
1706 l->rcv_unacked = 0;
1708 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1709 l->snd_nxt = l->rcv_nxt;
1710 return TIPC_LINK_SND_STATE;
1712 /* Unicast ACK */
1713 l->rcv_unacked = 0;
1714 l->stats.sent_acks++;
1715 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1716 return 0;
1719 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1721 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1723 int mtyp = RESET_MSG;
1724 struct sk_buff *skb;
1726 if (l->state == LINK_ESTABLISHING)
1727 mtyp = ACTIVATE_MSG;
1729 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1731 /* Inform peer that this endpoint is going down if applicable */
1732 skb = skb_peek_tail(xmitq);
1733 if (skb && (l->state == LINK_RESET))
1734 msg_set_peer_stopping(buf_msg(skb), 1);
1737 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1738 * Note that sending of broadcast NACK is coordinated among nodes, to
1739 * reduce the risk of NACK storms towards the sender
1741 static int tipc_link_build_nack_msg(struct tipc_link *l,
1742 struct sk_buff_head *xmitq)
1744 u32 def_cnt = ++l->stats.deferred_recv;
1745 struct sk_buff_head *dfq = &l->deferdq;
1746 u32 defq_len = skb_queue_len(dfq);
1747 int match1, match2;
1749 if (link_is_bc_rcvlink(l)) {
1750 match1 = def_cnt & 0xf;
1751 match2 = tipc_own_addr(l->net) & 0xf;
1752 if (match1 == match2)
1753 return TIPC_LINK_SND_STATE;
1754 return 0;
1757 if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1758 u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1760 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1761 rcvgap, 0, 0, xmitq);
1763 return 0;
1766 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1767 * @l: the link that should handle the message
1768 * @skb: TIPC packet
1769 * @xmitq: queue to place packets to be sent after this call
1771 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1772 struct sk_buff_head *xmitq)
1774 struct sk_buff_head *defq = &l->deferdq;
1775 struct tipc_msg *hdr = buf_msg(skb);
1776 u16 seqno, rcv_nxt, win_lim;
1777 int released = 0;
1778 int rc = 0;
1780 /* Verify and update link state */
1781 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1782 return tipc_link_proto_rcv(l, skb, xmitq);
1784 /* Don't send probe at next timeout expiration */
1785 l->silent_intv_cnt = 0;
1787 do {
1788 hdr = buf_msg(skb);
1789 seqno = msg_seqno(hdr);
1790 rcv_nxt = l->rcv_nxt;
1791 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1793 if (unlikely(!link_is_up(l))) {
1794 if (l->state == LINK_ESTABLISHING)
1795 rc = TIPC_LINK_UP_EVT;
1796 kfree_skb(skb);
1797 break;
1800 /* Drop if outside receive window */
1801 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1802 l->stats.duplicates++;
1803 kfree_skb(skb);
1804 break;
1806 released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1807 NULL, NULL, NULL, NULL);
1809 /* Defer delivery if sequence gap */
1810 if (unlikely(seqno != rcv_nxt)) {
1811 if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1812 l->stats.duplicates++;
1813 rc |= tipc_link_build_nack_msg(l, xmitq);
1814 break;
1817 /* Deliver packet */
1818 l->rcv_nxt++;
1819 l->stats.recv_pkts++;
1821 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1822 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1823 else if (!tipc_data_input(l, skb, l->inputq))
1824 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1825 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1826 rc |= tipc_link_build_state_msg(l, xmitq);
1827 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1828 break;
1829 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1831 /* Forward queues and wake up waiting users */
1832 if (released) {
1833 tipc_link_update_cwin(l, released, 0);
1834 tipc_link_advance_backlog(l, xmitq);
1835 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1836 link_prepare_wakeup(l);
1838 return rc;
1841 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1842 bool probe_reply, u16 rcvgap,
1843 int tolerance, int priority,
1844 struct sk_buff_head *xmitq)
1846 struct tipc_mon_state *mstate = &l->mon_state;
1847 struct sk_buff_head *dfq = &l->deferdq;
1848 struct tipc_link *bcl = l->bc_rcvlink;
1849 struct tipc_msg *hdr;
1850 struct sk_buff *skb;
1851 bool node_up = link_is_up(bcl);
1852 u16 glen = 0, bc_rcvgap = 0;
1853 int dlen = 0;
1854 void *data;
1856 /* Don't send protocol message during reset or link failover */
1857 if (tipc_link_is_blocked(l))
1858 return;
1860 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1861 return;
1863 if ((probe || probe_reply) && !skb_queue_empty(dfq))
1864 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1866 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1867 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1868 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1869 if (!skb)
1870 return;
1872 hdr = buf_msg(skb);
1873 data = msg_data(hdr);
1874 msg_set_session(hdr, l->session);
1875 msg_set_bearer_id(hdr, l->bearer_id);
1876 msg_set_net_plane(hdr, l->net_plane);
1877 msg_set_next_sent(hdr, l->snd_nxt);
1878 msg_set_ack(hdr, l->rcv_nxt - 1);
1879 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1880 msg_set_bc_ack_invalid(hdr, !node_up);
1881 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1882 msg_set_link_tolerance(hdr, tolerance);
1883 msg_set_linkprio(hdr, priority);
1884 msg_set_redundant_link(hdr, node_up);
1885 msg_set_seq_gap(hdr, 0);
1886 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1888 if (mtyp == STATE_MSG) {
1889 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1890 msg_set_seqno(hdr, l->snd_nxt_state++);
1891 msg_set_seq_gap(hdr, rcvgap);
1892 bc_rcvgap = link_bc_rcv_gap(bcl);
1893 msg_set_bc_gap(hdr, bc_rcvgap);
1894 msg_set_probe(hdr, probe);
1895 msg_set_is_keepalive(hdr, probe || probe_reply);
1896 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1897 glen = tipc_build_gap_ack_blks(l, hdr);
1898 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1899 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1900 skb_trim(skb, INT_H_SIZE + glen + dlen);
1901 l->stats.sent_states++;
1902 l->rcv_unacked = 0;
1903 } else {
1904 /* RESET_MSG or ACTIVATE_MSG */
1905 if (mtyp == ACTIVATE_MSG) {
1906 msg_set_dest_session_valid(hdr, 1);
1907 msg_set_dest_session(hdr, l->peer_session);
1909 msg_set_max_pkt(hdr, l->advertised_mtu);
1910 strcpy(data, l->if_name);
1911 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1912 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1914 if (probe)
1915 l->stats.sent_probes++;
1916 if (rcvgap)
1917 l->stats.sent_nacks++;
1918 if (bc_rcvgap)
1919 bcl->stats.sent_nacks++;
1920 skb->priority = TC_PRIO_CONTROL;
1921 __skb_queue_tail(xmitq, skb);
1922 trace_tipc_proto_build(skb, false, l->name);
1925 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1926 struct sk_buff_head *xmitq)
1928 u32 onode = tipc_own_addr(l->net);
1929 struct tipc_msg *hdr, *ihdr;
1930 struct sk_buff_head tnlq;
1931 struct sk_buff *skb;
1932 u32 dnode = l->addr;
1934 __skb_queue_head_init(&tnlq);
1935 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1936 INT_H_SIZE, BASIC_H_SIZE,
1937 dnode, onode, 0, 0, 0);
1938 if (!skb) {
1939 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1940 return;
1943 hdr = buf_msg(skb);
1944 msg_set_msgcnt(hdr, 1);
1945 msg_set_bearer_id(hdr, l->peer_bearer_id);
1947 ihdr = (struct tipc_msg *)msg_data(hdr);
1948 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1949 BASIC_H_SIZE, dnode);
1950 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1951 __skb_queue_tail(&tnlq, skb);
1952 tipc_link_xmit(l, &tnlq, xmitq);
1955 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1956 * with contents of the link's transmit and backlog queues.
1958 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1959 int mtyp, struct sk_buff_head *xmitq)
1961 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1962 struct sk_buff *skb, *tnlskb;
1963 struct tipc_msg *hdr, tnlhdr;
1964 struct sk_buff_head *queue = &l->transmq;
1965 struct sk_buff_head tmpxq, tnlq, frags;
1966 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1967 bool pktcnt_need_update = false;
1968 u16 syncpt;
1969 int rc;
1971 if (!tnl)
1972 return;
1974 __skb_queue_head_init(&tnlq);
1975 /* Link Synching:
1976 * From now on, send only one single ("dummy") SYNCH message
1977 * to peer. The SYNCH message does not contain any data, just
1978 * a header conveying the synch point to the peer.
1980 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1981 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1982 INT_H_SIZE, 0, l->addr,
1983 tipc_own_addr(l->net),
1984 0, 0, 0);
1985 if (!tnlskb) {
1986 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1987 link_co_err);
1988 return;
1991 hdr = buf_msg(tnlskb);
1992 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1993 msg_set_syncpt(hdr, syncpt);
1994 msg_set_bearer_id(hdr, l->peer_bearer_id);
1995 __skb_queue_tail(&tnlq, tnlskb);
1996 tipc_link_xmit(tnl, &tnlq, xmitq);
1997 return;
2000 __skb_queue_head_init(&tmpxq);
2001 __skb_queue_head_init(&frags);
2002 /* At least one packet required for safe algorithm => add dummy */
2003 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
2004 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
2005 0, 0, TIPC_ERR_NO_PORT);
2006 if (!skb) {
2007 pr_warn("%sunable to create tunnel packet\n", link_co_err);
2008 return;
2010 __skb_queue_tail(&tnlq, skb);
2011 tipc_link_xmit(l, &tnlq, &tmpxq);
2012 __skb_queue_purge(&tmpxq);
2014 /* Initialize reusable tunnel packet header */
2015 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
2016 mtyp, INT_H_SIZE, l->addr);
2017 if (mtyp == SYNCH_MSG)
2018 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
2019 else
2020 pktcnt = skb_queue_len(&l->transmq);
2021 pktcnt += skb_queue_len(&l->backlogq);
2022 msg_set_msgcnt(&tnlhdr, pktcnt);
2023 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
2024 tnl:
2025 /* Wrap each packet into a tunnel packet */
2026 skb_queue_walk(queue, skb) {
2027 hdr = buf_msg(skb);
2028 if (queue == &l->backlogq)
2029 msg_set_seqno(hdr, seqno++);
2030 pktlen = msg_size(hdr);
2032 /* Tunnel link MTU is not large enough? This could be
2033 * due to:
2034 * 1) Link MTU has just changed or set differently;
2035 * 2) Or FAILOVER on the top of a SYNCH message
2037 * The 2nd case should not happen if peer supports
2038 * TIPC_TUNNEL_ENHANCED
2040 if (pktlen > tnl->mtu - INT_H_SIZE) {
2041 if (mtyp == FAILOVER_MSG &&
2042 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2043 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2044 &frags);
2045 if (rc) {
2046 pr_warn("%sunable to frag msg: rc %d\n",
2047 link_co_err, rc);
2048 return;
2050 pktcnt += skb_queue_len(&frags) - 1;
2051 pktcnt_need_update = true;
2052 skb_queue_splice_tail_init(&frags, &tnlq);
2053 continue;
2055 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2056 * => Just warn it and return!
2058 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2059 link_co_err, msg_user(hdr),
2060 msg_type(hdr), msg_size(hdr));
2061 return;
2064 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2065 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2066 if (!tnlskb) {
2067 pr_warn("%sunable to send packet\n", link_co_err);
2068 return;
2070 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2071 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2072 __skb_queue_tail(&tnlq, tnlskb);
2074 if (queue != &l->backlogq) {
2075 queue = &l->backlogq;
2076 goto tnl;
2079 if (pktcnt_need_update)
2080 skb_queue_walk(&tnlq, skb) {
2081 hdr = buf_msg(skb);
2082 msg_set_msgcnt(hdr, pktcnt);
2085 tipc_link_xmit(tnl, &tnlq, xmitq);
2087 if (mtyp == FAILOVER_MSG) {
2088 tnl->drop_point = l->rcv_nxt;
2089 tnl->failover_reasm_skb = l->reasm_buf;
2090 l->reasm_buf = NULL;
2092 /* Failover the link's deferdq */
2093 if (unlikely(!skb_queue_empty(fdefq))) {
2094 pr_warn("Link failover deferdq not empty: %d!\n",
2095 skb_queue_len(fdefq));
2096 __skb_queue_purge(fdefq);
2098 skb_queue_splice_init(&l->deferdq, fdefq);
2103 * tipc_link_failover_prepare() - prepare tnl for link failover
2105 * This is a special version of the precursor - tipc_link_tnl_prepare(),
2106 * see the tipc_node_link_failover() for details
2108 * @l: failover link
2109 * @tnl: tunnel link
2110 * @xmitq: queue for messages to be xmited
2112 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2113 struct sk_buff_head *xmitq)
2115 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2117 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2119 /* This failover link endpoint was never established before,
2120 * so it has not received anything from peer.
2121 * Otherwise, it must be a normal failover situation or the
2122 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2123 * would have to start over from scratch instead.
2125 tnl->drop_point = 1;
2126 tnl->failover_reasm_skb = NULL;
2128 /* Initiate the link's failover deferdq */
2129 if (unlikely(!skb_queue_empty(fdefq))) {
2130 pr_warn("Link failover deferdq not empty: %d!\n",
2131 skb_queue_len(fdefq));
2132 __skb_queue_purge(fdefq);
2136 /* tipc_link_validate_msg(): validate message against current link state
2137 * Returns true if message should be accepted, otherwise false
2139 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2141 u16 curr_session = l->peer_session;
2142 u16 session = msg_session(hdr);
2143 int mtyp = msg_type(hdr);
2145 if (msg_user(hdr) != LINK_PROTOCOL)
2146 return true;
2148 switch (mtyp) {
2149 case RESET_MSG:
2150 if (!l->in_session)
2151 return true;
2152 /* Accept only RESET with new session number */
2153 return more(session, curr_session);
2154 case ACTIVATE_MSG:
2155 if (!l->in_session)
2156 return true;
2157 /* Accept only ACTIVATE with new or current session number */
2158 return !less(session, curr_session);
2159 case STATE_MSG:
2160 /* Accept only STATE with current session number */
2161 if (!l->in_session)
2162 return false;
2163 if (session != curr_session)
2164 return false;
2165 /* Extra sanity check */
2166 if (!link_is_up(l) && msg_ack(hdr))
2167 return false;
2168 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2169 return true;
2170 /* Accept only STATE with new sequence number */
2171 return !less(msg_seqno(hdr), l->rcv_nxt_state);
2172 default:
2173 return false;
2177 /* tipc_link_proto_rcv(): receive link level protocol message :
2178 * Note that network plane id propagates through the network, and may
2179 * change at any time. The node with lowest numerical id determines
2180 * network plane
2182 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2183 struct sk_buff_head *xmitq)
2185 struct tipc_msg *hdr = buf_msg(skb);
2186 struct tipc_gap_ack_blks *ga = NULL;
2187 bool reply = msg_probe(hdr), retransmitted = false;
2188 u16 dlen = msg_data_sz(hdr), glen = 0;
2189 u16 peers_snd_nxt = msg_next_sent(hdr);
2190 u16 peers_tol = msg_link_tolerance(hdr);
2191 u16 peers_prio = msg_linkprio(hdr);
2192 u16 gap = msg_seq_gap(hdr);
2193 u16 ack = msg_ack(hdr);
2194 u16 rcv_nxt = l->rcv_nxt;
2195 u16 rcvgap = 0;
2196 int mtyp = msg_type(hdr);
2197 int rc = 0, released;
2198 char *if_name;
2199 void *data;
2201 trace_tipc_proto_rcv(skb, false, l->name);
2202 if (tipc_link_is_blocked(l) || !xmitq)
2203 goto exit;
2205 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2206 l->net_plane = msg_net_plane(hdr);
2208 skb_linearize(skb);
2209 hdr = buf_msg(skb);
2210 data = msg_data(hdr);
2212 if (!tipc_link_validate_msg(l, hdr)) {
2213 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2214 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2215 goto exit;
2218 switch (mtyp) {
2219 case RESET_MSG:
2220 case ACTIVATE_MSG:
2221 /* Complete own link name with peer's interface name */
2222 if_name = strrchr(l->name, ':') + 1;
2223 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2224 break;
2225 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2226 break;
2227 strncpy(if_name, data, TIPC_MAX_IF_NAME);
2229 /* Update own tolerance if peer indicates a non-zero value */
2230 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2231 l->tolerance = peers_tol;
2232 l->bc_rcvlink->tolerance = peers_tol;
2234 /* Update own priority if peer's priority is higher */
2235 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2236 l->priority = peers_prio;
2238 /* If peer is going down we want full re-establish cycle */
2239 if (msg_peer_stopping(hdr)) {
2240 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2241 break;
2244 /* If this endpoint was re-created while peer was ESTABLISHING
2245 * it doesn't know current session number. Force re-synch.
2247 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2248 l->session != msg_dest_session(hdr)) {
2249 if (less(l->session, msg_dest_session(hdr)))
2250 l->session = msg_dest_session(hdr) + 1;
2251 break;
2254 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2255 if (mtyp == RESET_MSG || !link_is_up(l))
2256 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2258 /* ACTIVATE_MSG takes up link if it was already locally reset */
2259 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2260 rc = TIPC_LINK_UP_EVT;
2262 l->peer_session = msg_session(hdr);
2263 l->in_session = true;
2264 l->peer_bearer_id = msg_bearer_id(hdr);
2265 if (l->mtu > msg_max_pkt(hdr))
2266 l->mtu = msg_max_pkt(hdr);
2267 break;
2269 case STATE_MSG:
2270 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2272 /* Update own tolerance if peer indicates a non-zero value */
2273 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2274 l->tolerance = peers_tol;
2275 l->bc_rcvlink->tolerance = peers_tol;
2277 /* Update own prio if peer indicates a different value */
2278 if ((peers_prio != l->priority) &&
2279 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2280 l->priority = peers_prio;
2281 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2284 l->silent_intv_cnt = 0;
2285 l->stats.recv_states++;
2286 if (msg_probe(hdr))
2287 l->stats.recv_probes++;
2289 if (!link_is_up(l)) {
2290 if (l->state == LINK_ESTABLISHING)
2291 rc = TIPC_LINK_UP_EVT;
2292 break;
2295 /* Receive Gap ACK blocks from peer if any */
2296 glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2298 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2299 &l->mon_state, l->bearer_id);
2301 /* Send NACK if peer has sent pkts we haven't received yet */
2302 if ((reply || msg_is_keepalive(hdr)) &&
2303 more(peers_snd_nxt, rcv_nxt) &&
2304 !tipc_link_is_synching(l) &&
2305 skb_queue_empty(&l->deferdq))
2306 rcvgap = peers_snd_nxt - l->rcv_nxt;
2307 if (rcvgap || reply)
2308 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2309 rcvgap, 0, 0, xmitq);
2311 released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2312 &retransmitted, &rc);
2313 if (gap)
2314 l->stats.recv_nacks++;
2315 if (released || retransmitted)
2316 tipc_link_update_cwin(l, released, retransmitted);
2317 if (released)
2318 tipc_link_advance_backlog(l, xmitq);
2319 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2320 link_prepare_wakeup(l);
2322 exit:
2323 kfree_skb(skb);
2324 return rc;
2327 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2329 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2330 u16 peers_snd_nxt,
2331 struct sk_buff_head *xmitq)
2333 struct sk_buff *skb;
2334 struct tipc_msg *hdr;
2335 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2336 u16 ack = l->rcv_nxt - 1;
2337 u16 gap_to = peers_snd_nxt - 1;
2339 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2340 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2341 if (!skb)
2342 return false;
2343 hdr = buf_msg(skb);
2344 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2345 msg_set_bcast_ack(hdr, ack);
2346 msg_set_bcgap_after(hdr, ack);
2347 if (dfrd_skb)
2348 gap_to = buf_seqno(dfrd_skb) - 1;
2349 msg_set_bcgap_to(hdr, gap_to);
2350 msg_set_non_seq(hdr, bcast);
2351 __skb_queue_tail(xmitq, skb);
2352 return true;
2355 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2357 * Give a newly added peer node the sequence number where it should
2358 * start receiving and acking broadcast packets.
2360 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2361 struct sk_buff_head *xmitq)
2363 struct sk_buff_head list;
2365 __skb_queue_head_init(&list);
2366 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2367 return;
2368 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2369 tipc_link_xmit(l, &list, xmitq);
2372 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2374 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2376 int mtyp = msg_type(hdr);
2377 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2379 if (link_is_up(l))
2380 return;
2382 if (msg_user(hdr) == BCAST_PROTOCOL) {
2383 l->rcv_nxt = peers_snd_nxt;
2384 l->state = LINK_ESTABLISHED;
2385 return;
2388 if (l->peer_caps & TIPC_BCAST_SYNCH)
2389 return;
2391 if (msg_peer_node_is_up(hdr))
2392 return;
2394 /* Compatibility: accept older, less safe initial synch data */
2395 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2396 l->rcv_nxt = peers_snd_nxt;
2399 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2401 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2402 struct sk_buff_head *xmitq)
2404 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2405 int rc = 0;
2407 if (!link_is_up(l))
2408 return rc;
2410 if (!msg_peer_node_is_up(hdr))
2411 return rc;
2413 /* Open when peer acknowledges our bcast init msg (pkt #1) */
2414 if (msg_ack(hdr))
2415 l->bc_peer_is_up = true;
2417 if (!l->bc_peer_is_up)
2418 return rc;
2420 /* Ignore if peers_snd_nxt goes beyond receive window */
2421 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2422 return rc;
2424 l->snd_nxt = peers_snd_nxt;
2425 if (link_bc_rcv_gap(l))
2426 rc |= TIPC_LINK_SND_STATE;
2428 /* Return now if sender supports nack via STATE messages */
2429 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2430 return rc;
2432 /* Otherwise, be backwards compatible */
2434 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2435 l->nack_state = BC_NACK_SND_CONDITIONAL;
2436 return 0;
2439 /* Don't NACK if one was recently sent or peeked */
2440 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2441 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2442 return 0;
2445 /* Conditionally delay NACK sending until next synch rcv */
2446 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2447 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2448 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2449 return 0;
2452 /* Send NACK now but suppress next one */
2453 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2454 l->nack_state = BC_NACK_SND_SUPPRESS;
2455 return 0;
2458 int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2459 struct tipc_gap_ack_blks *ga,
2460 struct sk_buff_head *xmitq,
2461 struct sk_buff_head *retrq)
2463 struct tipc_link *l = r->bc_sndlink;
2464 bool unused = false;
2465 int rc = 0;
2467 if (!link_is_up(r) || !r->bc_peer_is_up)
2468 return 0;
2470 if (gap) {
2471 l->stats.recv_nacks++;
2472 r->stats.recv_nacks++;
2475 if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2476 return 0;
2478 trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2479 tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2481 tipc_link_advance_backlog(l, xmitq);
2482 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2483 link_prepare_wakeup(l);
2485 return rc;
2488 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2489 * This function is here for backwards compatibility, since
2490 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2492 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2493 struct sk_buff_head *xmitq)
2495 struct tipc_msg *hdr = buf_msg(skb);
2496 u32 dnode = msg_destnode(hdr);
2497 int mtyp = msg_type(hdr);
2498 u16 acked = msg_bcast_ack(hdr);
2499 u16 from = acked + 1;
2500 u16 to = msg_bcgap_to(hdr);
2501 u16 peers_snd_nxt = to + 1;
2502 int rc = 0;
2504 kfree_skb(skb);
2506 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2507 return 0;
2509 if (mtyp != STATE_MSG)
2510 return 0;
2512 if (dnode == tipc_own_addr(l->net)) {
2513 rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2514 xmitq);
2515 l->stats.recv_nacks++;
2516 return rc;
2519 /* Msg for other node => suppress own NACK at next sync if applicable */
2520 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2521 l->nack_state = BC_NACK_SND_SUPPRESS;
2523 return 0;
2526 void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2528 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2530 l->min_win = min_win;
2531 l->ssthresh = max_win;
2532 l->max_win = max_win;
2533 l->window = min_win;
2534 l->backlog[TIPC_LOW_IMPORTANCE].limit = min_win * 2;
2535 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = min_win * 4;
2536 l->backlog[TIPC_HIGH_IMPORTANCE].limit = min_win * 6;
2537 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2538 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2542 * link_reset_stats - reset link statistics
2543 * @l: pointer to link
2545 void tipc_link_reset_stats(struct tipc_link *l)
2547 memset(&l->stats, 0, sizeof(l->stats));
2550 static void link_print(struct tipc_link *l, const char *str)
2552 struct sk_buff *hskb = skb_peek(&l->transmq);
2553 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2554 u16 tail = l->snd_nxt - 1;
2556 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2557 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2558 skb_queue_len(&l->transmq), head, tail,
2559 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2562 /* Parse and validate nested (link) properties valid for media, bearer and link
2564 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2566 int err;
2568 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2569 tipc_nl_prop_policy, NULL);
2570 if (err)
2571 return err;
2573 if (props[TIPC_NLA_PROP_PRIO]) {
2574 u32 prio;
2576 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2577 if (prio > TIPC_MAX_LINK_PRI)
2578 return -EINVAL;
2581 if (props[TIPC_NLA_PROP_TOL]) {
2582 u32 tol;
2584 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2585 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2586 return -EINVAL;
2589 if (props[TIPC_NLA_PROP_WIN]) {
2590 u32 max_win;
2592 max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2593 if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2594 return -EINVAL;
2597 return 0;
2600 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2602 int i;
2603 struct nlattr *stats;
2605 struct nla_map {
2606 u32 key;
2607 u32 val;
2610 struct nla_map map[] = {
2611 {TIPC_NLA_STATS_RX_INFO, 0},
2612 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2613 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2614 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2615 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2616 {TIPC_NLA_STATS_TX_INFO, 0},
2617 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2618 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2619 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2620 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2621 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2622 s->msg_length_counts : 1},
2623 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2624 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2625 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2626 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2627 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2628 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2629 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2630 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2631 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2632 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2633 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2634 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2635 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2636 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2637 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2638 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2639 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2640 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2641 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2642 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2643 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2644 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2645 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2648 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2649 if (!stats)
2650 return -EMSGSIZE;
2652 for (i = 0; i < ARRAY_SIZE(map); i++)
2653 if (nla_put_u32(skb, map[i].key, map[i].val))
2654 goto msg_full;
2656 nla_nest_end(skb, stats);
2658 return 0;
2659 msg_full:
2660 nla_nest_cancel(skb, stats);
2662 return -EMSGSIZE;
2665 /* Caller should hold appropriate locks to protect the link */
2666 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2667 struct tipc_link *link, int nlflags)
2669 u32 self = tipc_own_addr(net);
2670 struct nlattr *attrs;
2671 struct nlattr *prop;
2672 void *hdr;
2673 int err;
2675 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2676 nlflags, TIPC_NL_LINK_GET);
2677 if (!hdr)
2678 return -EMSGSIZE;
2680 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2681 if (!attrs)
2682 goto msg_full;
2684 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2685 goto attr_msg_full;
2686 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2687 goto attr_msg_full;
2688 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2689 goto attr_msg_full;
2690 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2691 goto attr_msg_full;
2692 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2693 goto attr_msg_full;
2695 if (tipc_link_is_up(link))
2696 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2697 goto attr_msg_full;
2698 if (link->active)
2699 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2700 goto attr_msg_full;
2702 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2703 if (!prop)
2704 goto attr_msg_full;
2705 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2706 goto prop_msg_full;
2707 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2708 goto prop_msg_full;
2709 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2710 link->window))
2711 goto prop_msg_full;
2712 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2713 goto prop_msg_full;
2714 nla_nest_end(msg->skb, prop);
2716 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2717 if (err)
2718 goto attr_msg_full;
2720 nla_nest_end(msg->skb, attrs);
2721 genlmsg_end(msg->skb, hdr);
2723 return 0;
2725 prop_msg_full:
2726 nla_nest_cancel(msg->skb, prop);
2727 attr_msg_full:
2728 nla_nest_cancel(msg->skb, attrs);
2729 msg_full:
2730 genlmsg_cancel(msg->skb, hdr);
2732 return -EMSGSIZE;
2735 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2736 struct tipc_stats *stats)
2738 int i;
2739 struct nlattr *nest;
2741 struct nla_map {
2742 __u32 key;
2743 __u32 val;
2746 struct nla_map map[] = {
2747 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2748 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2749 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2750 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2751 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2752 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2753 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2754 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2755 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2756 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2757 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2758 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2759 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2760 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2761 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2762 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2763 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2764 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2765 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2766 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2769 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2770 if (!nest)
2771 return -EMSGSIZE;
2773 for (i = 0; i < ARRAY_SIZE(map); i++)
2774 if (nla_put_u32(skb, map[i].key, map[i].val))
2775 goto msg_full;
2777 nla_nest_end(skb, nest);
2779 return 0;
2780 msg_full:
2781 nla_nest_cancel(skb, nest);
2783 return -EMSGSIZE;
2786 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2787 struct tipc_link *bcl)
2789 int err;
2790 void *hdr;
2791 struct nlattr *attrs;
2792 struct nlattr *prop;
2793 u32 bc_mode = tipc_bcast_get_mode(net);
2794 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2796 if (!bcl)
2797 return 0;
2799 tipc_bcast_lock(net);
2801 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2802 NLM_F_MULTI, TIPC_NL_LINK_GET);
2803 if (!hdr) {
2804 tipc_bcast_unlock(net);
2805 return -EMSGSIZE;
2808 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2809 if (!attrs)
2810 goto msg_full;
2812 /* The broadcast link is always up */
2813 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2814 goto attr_msg_full;
2816 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2817 goto attr_msg_full;
2818 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2819 goto attr_msg_full;
2820 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2821 goto attr_msg_full;
2822 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2823 goto attr_msg_full;
2825 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2826 if (!prop)
2827 goto attr_msg_full;
2828 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2829 goto prop_msg_full;
2830 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2831 goto prop_msg_full;
2832 if (bc_mode & BCLINK_MODE_SEL)
2833 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2834 bc_ratio))
2835 goto prop_msg_full;
2836 nla_nest_end(msg->skb, prop);
2838 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2839 if (err)
2840 goto attr_msg_full;
2842 tipc_bcast_unlock(net);
2843 nla_nest_end(msg->skb, attrs);
2844 genlmsg_end(msg->skb, hdr);
2846 return 0;
2848 prop_msg_full:
2849 nla_nest_cancel(msg->skb, prop);
2850 attr_msg_full:
2851 nla_nest_cancel(msg->skb, attrs);
2852 msg_full:
2853 tipc_bcast_unlock(net);
2854 genlmsg_cancel(msg->skb, hdr);
2856 return -EMSGSIZE;
2859 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2860 struct sk_buff_head *xmitq)
2862 l->tolerance = tol;
2863 if (l->bc_rcvlink)
2864 l->bc_rcvlink->tolerance = tol;
2865 if (link_is_up(l))
2866 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2869 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2870 struct sk_buff_head *xmitq)
2872 l->priority = prio;
2873 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2876 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2878 l->abort_limit = limit;
2882 * tipc_link_dump - dump TIPC link data
2883 * @l: tipc link to be dumped
2884 * @dqueues: bitmask to decide if any link queue to be dumped?
2885 * - TIPC_DUMP_NONE: don't dump link queues
2886 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2887 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2888 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2889 * - TIPC_DUMP_INPUTQ: dump link input queue
2890 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2891 * - TIPC_DUMP_ALL: dump all the link queues above
2892 * @buf: returned buffer of dump data in format
2894 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2896 int i = 0;
2897 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2898 struct sk_buff_head *list;
2899 struct sk_buff *hskb, *tskb;
2900 u32 len;
2902 if (!l) {
2903 i += scnprintf(buf, sz, "link data: (null)\n");
2904 return i;
2907 i += scnprintf(buf, sz, "link data: %x", l->addr);
2908 i += scnprintf(buf + i, sz - i, " %x", l->state);
2909 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2910 i += scnprintf(buf + i, sz - i, " %u", l->session);
2911 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2912 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2913 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2914 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2915 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2916 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2917 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2918 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2919 i += scnprintf(buf + i, sz - i, " %u", 0);
2920 i += scnprintf(buf + i, sz - i, " %u", 0);
2921 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2923 list = &l->transmq;
2924 len = skb_queue_len(list);
2925 hskb = skb_peek(list);
2926 tskb = skb_peek_tail(list);
2927 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2928 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2929 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2931 list = &l->deferdq;
2932 len = skb_queue_len(list);
2933 hskb = skb_peek(list);
2934 tskb = skb_peek_tail(list);
2935 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2936 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2937 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2939 list = &l->backlogq;
2940 len = skb_queue_len(list);
2941 hskb = skb_peek(list);
2942 tskb = skb_peek_tail(list);
2943 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2944 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2945 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2947 list = l->inputq;
2948 len = skb_queue_len(list);
2949 hskb = skb_peek(list);
2950 tskb = skb_peek_tail(list);
2951 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2952 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2953 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2955 if (dqueues & TIPC_DUMP_TRANSMQ) {
2956 i += scnprintf(buf + i, sz - i, "transmq: ");
2957 i += tipc_list_dump(&l->transmq, false, buf + i);
2959 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2960 i += scnprintf(buf + i, sz - i,
2961 "backlogq: <%u %u %u %u %u>, ",
2962 l->backlog[TIPC_LOW_IMPORTANCE].len,
2963 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2964 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2965 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2966 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2967 i += tipc_list_dump(&l->backlogq, false, buf + i);
2969 if (dqueues & TIPC_DUMP_DEFERDQ) {
2970 i += scnprintf(buf + i, sz - i, "deferdq: ");
2971 i += tipc_list_dump(&l->deferdq, false, buf + i);
2973 if (dqueues & TIPC_DUMP_INPUTQ) {
2974 i += scnprintf(buf + i, sz - i, "inputq: ");
2975 i += tipc_list_dump(l->inputq, false, buf + i);
2977 if (dqueues & TIPC_DUMP_WAKEUP) {
2978 i += scnprintf(buf + i, sz - i, "wakeup: ");
2979 i += tipc_list_dump(&l->wakeupq, false, buf + i);
2982 return i;