Loading...
1/*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
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.
19 *
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.
23 *
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.
35 */
36
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"
48
49#include <linux/pkt_sched.h>
50
51struct 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 */
79};
80
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
151 */
152struct tipc_link {
153 u32 addr;
154 char name[TIPC_MAX_LINK_NAME];
155 struct net *net;
156
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;
176
177 /* Failover/synch */
178 u16 drop_point;
179 struct sk_buff *failover_reasm_skb;
180 struct sk_buff_head failover_deferdq;
181
182 /* Max packet negotiation */
183 u16 mtu;
184 u16 advertised_mtu;
185
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;
195
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;
202
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;
211
212 /* Fragmentation/reassembly */
213 struct sk_buff *reasm_buf;
214 struct sk_buff *reasm_tnlmsg;
215
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;
225
226 /* Statistics */
227 struct tipc_stats stats;
228};
229
230/*
231 * Error message prefixes
232 */
233static const char *link_co_err = "Link tunneling error, ";
234static const char *link_rst_msg = "Resetting link ";
235
236/* Send states for broadcast NACKs
237 */
238enum {
239 BC_NACK_SND_CONDITIONAL,
240 BC_NACK_SND_UNCONDITIONAL,
241 BC_NACK_SND_SUPPRESS,
242};
243
244#define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
245#define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
246
247/* Link FSM states:
248 */
249enum {
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
257};
258
259/* Link FSM state checking routines
260 */
261static int link_is_up(struct tipc_link *l)
262{
263 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
264}
265
266static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
267 struct sk_buff_head *xmitq);
268static 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);
272static void link_print(struct tipc_link *l, const char *str);
273static int tipc_link_build_nack_msg(struct tipc_link *l,
274 struct sk_buff_head *xmitq);
275static void tipc_link_build_bc_init_msg(struct tipc_link *l,
276 struct sk_buff_head *xmitq);
277static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
278 struct tipc_link *l, u8 start_index);
279static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr);
280static 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);
285static void tipc_link_update_cwin(struct tipc_link *l, int released,
286 bool retransmitted);
287/*
288 * Simple non-static link routines (i.e. referenced outside this file)
289 */
290bool tipc_link_is_up(struct tipc_link *l)
291{
292 return link_is_up(l);
293}
294
295bool tipc_link_peer_is_down(struct tipc_link *l)
296{
297 return l->state == LINK_PEER_RESET;
298}
299
300bool tipc_link_is_reset(struct tipc_link *l)
301{
302 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
303}
304
305bool tipc_link_is_establishing(struct tipc_link *l)
306{
307 return l->state == LINK_ESTABLISHING;
308}
309
310bool tipc_link_is_synching(struct tipc_link *l)
311{
312 return l->state == LINK_SYNCHING;
313}
314
315bool tipc_link_is_failingover(struct tipc_link *l)
316{
317 return l->state == LINK_FAILINGOVER;
318}
319
320bool tipc_link_is_blocked(struct tipc_link *l)
321{
322 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
323}
324
325static bool link_is_bc_sndlink(struct tipc_link *l)
326{
327 return !l->bc_sndlink;
328}
329
330static bool link_is_bc_rcvlink(struct tipc_link *l)
331{
332 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
333}
334
335void tipc_link_set_active(struct tipc_link *l, bool active)
336{
337 l->active = active;
338}
339
340u32 tipc_link_id(struct tipc_link *l)
341{
342 return l->peer_bearer_id << 16 | l->bearer_id;
343}
344
345int tipc_link_min_win(struct tipc_link *l)
346{
347 return l->min_win;
348}
349
350int tipc_link_max_win(struct tipc_link *l)
351{
352 return l->max_win;
353}
354
355int tipc_link_prio(struct tipc_link *l)
356{
357 return l->priority;
358}
359
360unsigned long tipc_link_tolerance(struct tipc_link *l)
361{
362 return l->tolerance;
363}
364
365struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
366{
367 return l->inputq;
368}
369
370char tipc_link_plane(struct tipc_link *l)
371{
372 return l->net_plane;
373}
374
375struct net *tipc_link_net(struct tipc_link *l)
376{
377 return l->net;
378}
379
380void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
381{
382 l->peer_caps = capabilities;
383}
384
385void tipc_link_add_bc_peer(struct tipc_link *snd_l,
386 struct tipc_link *uc_l,
387 struct sk_buff_head *xmitq)
388{
389 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
390
391 snd_l->ackers++;
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);
395}
396
397void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
398 struct tipc_link *rcv_l,
399 struct sk_buff_head *xmitq)
400{
401 u16 ack = snd_l->snd_nxt - 1;
402
403 snd_l->ackers--;
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);
415 }
416}
417
418int tipc_link_bc_peers(struct tipc_link *l)
419{
420 return l->ackers;
421}
422
423static u16 link_bc_rcv_gap(struct tipc_link *l)
424{
425 struct sk_buff *skb = skb_peek(&l->deferdq);
426 u16 gap = 0;
427
428 if (more(l->snd_nxt, l->rcv_nxt))
429 gap = l->snd_nxt - l->rcv_nxt;
430 if (skb)
431 gap = buf_seqno(skb) - l->rcv_nxt;
432 return gap;
433}
434
435void tipc_link_set_mtu(struct tipc_link *l, int mtu)
436{
437 l->mtu = mtu;
438}
439
440int tipc_link_mtu(struct tipc_link *l)
441{
442 return l->mtu;
443}
444
445int tipc_link_mss(struct tipc_link *l)
446{
447#ifdef CONFIG_TIPC_CRYPTO
448 return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
449#else
450 return l->mtu - INT_H_SIZE;
451#endif
452}
453
454u16 tipc_link_rcv_nxt(struct tipc_link *l)
455{
456 return l->rcv_nxt;
457}
458
459u16 tipc_link_acked(struct tipc_link *l)
460{
461 return l->acked;
462}
463
464char *tipc_link_name(struct tipc_link *l)
465{
466 return l->name;
467}
468
469u32 tipc_link_state(struct tipc_link *l)
470{
471 return l->state;
472}
473
474/**
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
495 *
496 * Return: true if link was created, otherwise false
497 */
498bool 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)
507{
508 char peer_str[NODE_ID_STR_LEN] = {0,};
509 char self_str[NODE_ID_STR_LEN] = {0,};
510 struct tipc_link *l;
511
512 l = kzalloc(sizeof(*l), GFP_ATOMIC);
513 if (!l)
514 return false;
515 *link = l;
516 l->session = session;
517
518 /* Set link name for unicast links only */
519 if (peer_id) {
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);
526 }
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);
530
531 strcpy(l->if_name, if_name);
532 l->addr = peer;
533 l->peer_caps = peer_caps;
534 l->net = net;
535 l->in_session = false;
536 l->bearer_id = bearer_id;
537 l->tolerance = tolerance;
538 if (bc_rcvlink)
539 bc_rcvlink->tolerance = tolerance;
540 l->net_plane = net_plane;
541 l->advertised_mtu = mtu;
542 l->mtu = mtu;
543 l->priority = priority;
544 tipc_link_set_queue_limits(l, min_win, max_win);
545 l->ackers = 1;
546 l->bc_sndlink = bc_sndlink;
547 l->bc_rcvlink = bc_rcvlink;
548 l->inputq = inputq;
549 l->namedq = namedq;
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);
557 return true;
558}
559
560/**
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
574 *
575 * Return: true if link was created, otherwise false
576 */
577bool 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)
583{
584 struct tipc_link *l;
585
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))
589 return false;
590
591 l = *link;
592 if (peer_id) {
593 char peer_str[NODE_ID_STR_LEN] = {0,};
594
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,
600 peer_str);
601 } else {
602 strcpy(l->name, tipc_bclink_name);
603 }
604 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
605 tipc_link_reset(l);
606 l->state = LINK_RESET;
607 l->ackers = 0;
608 l->bc_rcvlink = l;
609
610 /* Broadcast send link is always up */
611 if (link_is_bc_sndlink(l))
612 l->state = LINK_ESTABLISHED;
613
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);
617
618 return true;
619}
620
621/**
622 * tipc_link_fsm_evt - link finite state machine
623 * @l: pointer to link
624 * @evt: state machine event to be processed
625 */
626int tipc_link_fsm_evt(struct tipc_link *l, int evt)
627{
628 int rc = 0;
629 int old_state = l->state;
630
631 switch (l->state) {
632 case LINK_RESETTING:
633 switch (evt) {
634 case LINK_PEER_RESET_EVT:
635 l->state = LINK_PEER_RESET;
636 break;
637 case LINK_RESET_EVT:
638 l->state = LINK_RESET;
639 break;
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:
646 default:
647 goto illegal_evt;
648 }
649 break;
650 case LINK_RESET:
651 switch (evt) {
652 case LINK_PEER_RESET_EVT:
653 l->state = LINK_ESTABLISHING;
654 break;
655 case LINK_FAILOVER_BEGIN_EVT:
656 l->state = LINK_FAILINGOVER;
657 break;
658 case LINK_FAILURE_EVT:
659 case LINK_RESET_EVT:
660 case LINK_ESTABLISH_EVT:
661 case LINK_FAILOVER_END_EVT:
662 break;
663 case LINK_SYNCH_BEGIN_EVT:
664 case LINK_SYNCH_END_EVT:
665 default:
666 goto illegal_evt;
667 }
668 break;
669 case LINK_PEER_RESET:
670 switch (evt) {
671 case LINK_RESET_EVT:
672 l->state = LINK_ESTABLISHING;
673 break;
674 case LINK_PEER_RESET_EVT:
675 case LINK_ESTABLISH_EVT:
676 case LINK_FAILURE_EVT:
677 break;
678 case LINK_SYNCH_BEGIN_EVT:
679 case LINK_SYNCH_END_EVT:
680 case LINK_FAILOVER_BEGIN_EVT:
681 case LINK_FAILOVER_END_EVT:
682 default:
683 goto illegal_evt;
684 }
685 break;
686 case LINK_FAILINGOVER:
687 switch (evt) {
688 case LINK_FAILOVER_END_EVT:
689 l->state = LINK_RESET;
690 break;
691 case LINK_PEER_RESET_EVT:
692 case LINK_RESET_EVT:
693 case LINK_ESTABLISH_EVT:
694 case LINK_FAILURE_EVT:
695 break;
696 case LINK_FAILOVER_BEGIN_EVT:
697 case LINK_SYNCH_BEGIN_EVT:
698 case LINK_SYNCH_END_EVT:
699 default:
700 goto illegal_evt;
701 }
702 break;
703 case LINK_ESTABLISHING:
704 switch (evt) {
705 case LINK_ESTABLISH_EVT:
706 l->state = LINK_ESTABLISHED;
707 break;
708 case LINK_FAILOVER_BEGIN_EVT:
709 l->state = LINK_FAILINGOVER;
710 break;
711 case LINK_RESET_EVT:
712 l->state = LINK_RESET;
713 break;
714 case LINK_FAILURE_EVT:
715 case LINK_PEER_RESET_EVT:
716 case LINK_SYNCH_BEGIN_EVT:
717 case LINK_FAILOVER_END_EVT:
718 break;
719 case LINK_SYNCH_END_EVT:
720 default:
721 goto illegal_evt;
722 }
723 break;
724 case LINK_ESTABLISHED:
725 switch (evt) {
726 case LINK_PEER_RESET_EVT:
727 l->state = LINK_PEER_RESET;
728 rc |= TIPC_LINK_DOWN_EVT;
729 break;
730 case LINK_FAILURE_EVT:
731 l->state = LINK_RESETTING;
732 rc |= TIPC_LINK_DOWN_EVT;
733 break;
734 case LINK_RESET_EVT:
735 l->state = LINK_RESET;
736 break;
737 case LINK_ESTABLISH_EVT:
738 case LINK_SYNCH_END_EVT:
739 break;
740 case LINK_SYNCH_BEGIN_EVT:
741 l->state = LINK_SYNCHING;
742 break;
743 case LINK_FAILOVER_BEGIN_EVT:
744 case LINK_FAILOVER_END_EVT:
745 default:
746 goto illegal_evt;
747 }
748 break;
749 case LINK_SYNCHING:
750 switch (evt) {
751 case LINK_PEER_RESET_EVT:
752 l->state = LINK_PEER_RESET;
753 rc |= TIPC_LINK_DOWN_EVT;
754 break;
755 case LINK_FAILURE_EVT:
756 l->state = LINK_RESETTING;
757 rc |= TIPC_LINK_DOWN_EVT;
758 break;
759 case LINK_RESET_EVT:
760 l->state = LINK_RESET;
761 break;
762 case LINK_ESTABLISH_EVT:
763 case LINK_SYNCH_BEGIN_EVT:
764 break;
765 case LINK_SYNCH_END_EVT:
766 l->state = LINK_ESTABLISHED;
767 break;
768 case LINK_FAILOVER_BEGIN_EVT:
769 case LINK_FAILOVER_END_EVT:
770 default:
771 goto illegal_evt;
772 }
773 break;
774 default:
775 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
776 }
777 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
778 return rc;
779illegal_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);
783 return rc;
784}
785
786/* link_profile_stats - update statistical profiling of traffic
787 */
788static void link_profile_stats(struct tipc_link *l)
789{
790 struct sk_buff *skb;
791 struct tipc_msg *msg;
792 int length;
793
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++;
797
798 skb = skb_peek(&l->transmq);
799 if (!skb)
800 return;
801 msg = buf_msg(skb);
802 length = msg_size(msg);
803
804 if (msg_user(msg) == MSG_FRAGMENTER) {
805 if (msg_type(msg) != FIRST_FRAGMENT)
806 return;
807 length = msg_size(msg_inner_hdr(msg));
808 }
809 l->stats.msg_lengths_total += length;
810 l->stats.msg_length_counts++;
811 if (length <= 64)
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]++;
823 else
824 l->stats.msg_length_profile[6]++;
825}
826
827/**
828 * tipc_link_too_silent - check if link is "too silent"
829 * @l: tipc link to be checked
830 *
831 * Return: true if the link 'silent_intv_cnt' is about to reach the
832 * 'abort_limit' value, otherwise false
833 */
834bool tipc_link_too_silent(struct tipc_link *l)
835{
836 return (l->silent_intv_cnt + 2 > l->abort_limit);
837}
838
839/* tipc_link_timeout - perform periodic task as instructed from node timeout
840 */
841int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
842{
843 int mtyp = 0;
844 int rc = 0;
845 bool state = false;
846 bool probe = false;
847 bool setup = false;
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;
851
852 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
853 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
854 switch (l->state) {
855 case LINK_ESTABLISHED:
856 case LINK_SYNCHING:
857 mtyp = STATE_MSG;
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);
873 probe = true;
874 }
875 l->checkpoint = l->snd_nxt;
876 break;
877 case LINK_RESET:
878 setup = l->rst_cnt++ <= 4;
879 setup |= !(l->rst_cnt % 16);
880 mtyp = RESET_MSG;
881 break;
882 case LINK_ESTABLISHING:
883 setup = true;
884 mtyp = ACTIVATE_MSG;
885 break;
886 case LINK_PEER_RESET:
887 case LINK_RESETTING:
888 case LINK_FAILINGOVER:
889 break;
890 default:
891 break;
892 }
893
894 if (state || probe || setup)
895 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
896
897 return rc;
898}
899
900/**
901 * link_schedule_user - schedule a message sender for wakeup after congestion
902 * @l: congested link
903 * @hdr: header of message that is being sent
904 * Create pseudo msg to send back to user when congestion abates
905 */
906static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
907{
908 u32 dnode = tipc_own_addr(l->net);
909 u32 dport = msg_origport(hdr);
910 struct sk_buff *skb;
911
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);
915 if (!skb)
916 return -ENOBUFS;
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!");
922 return -ELINKCONG;
923}
924
925/**
926 * link_prepare_wakeup - prepare users for wakeup after congestion
927 * @l: congested link
928 * Wake up a number of waiting users, as permitted by available space
929 * in the send queue
930 */
931static void link_prepare_wakeup(struct tipc_link *l)
932{
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;
937 int avail[5] = {0,};
938 int imp = 0;
939
940 __skb_queue_head_init(&tmpq);
941
942 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
943 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
944
945 skb_queue_walk_safe(wakeupq, skb, tmp) {
946 imp = TIPC_SKB_CB(skb)->chain_imp;
947 if (avail[imp] <= 0)
948 continue;
949 avail[imp]--;
950 __skb_unlink(skb, wakeupq);
951 __skb_queue_tail(&tmpq, skb);
952 }
953
954 spin_lock_bh(&inputq->lock);
955 skb_queue_splice_tail(&tmpq, inputq);
956 spin_unlock_bh(&inputq->lock);
957
958}
959
960/**
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
965 */
966static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
967 struct tipc_link *l)
968{
969 if (link_is_bc_sndlink(l))
970 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
971 else
972 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
973}
974
975void tipc_link_reset(struct tipc_link *l)
976{
977 struct sk_buff_head list;
978 u32 imp;
979
980 __skb_queue_head_init(&list);
981
982 l->in_session = false;
983 /* Force re-synch of peer session number before establishing */
984 l->peer_session--;
985 l->session++;
986 l->mtu = l->advertised_mtu;
987
988 spin_lock_bh(&l->wakeupq.lock);
989 skb_queue_splice_init(&l->wakeupq, &list);
990 spin_unlock_bh(&l->wakeupq.lock);
991
992 spin_lock_bh(&l->inputq->lock);
993 skb_queue_splice_init(&list, l->inputq);
994 spin_unlock_bh(&l->inputq->lock);
995
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;
1003 }
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;
1010 l->rcv_unacked = 0;
1011 l->snd_nxt = 1;
1012 l->rcv_nxt = 1;
1013 l->snd_nxt_state = 1;
1014 l->rcv_nxt_state = 1;
1015 l->acked = 0;
1016 l->last_gap = 0;
1017 kfree(l->last_ga);
1018 l->last_ga = NULL;
1019 l->silent_intv_cnt = 0;
1020 l->rst_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);
1024}
1025
1026/**
1027 * tipc_link_xmit(): enqueue buffer list according to queue situation
1028 * @l: link to use
1029 * @list: chain of buffers containing message
1030 * @xmitq: returned list of packets to be sent by caller
1031 *
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
1035 */
1036int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
1037 struct sk_buff_head *xmitq)
1038{
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;
1050 bool new_bundle;
1051 int rc = 0;
1052 int imp;
1053
1054 if (pkt_cnt <= 0)
1055 return 0;
1056
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);
1063 return -EMSGSIZE;
1064 }
1065
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);
1071 return -ENOBUFS;
1072 }
1073 rc = link_schedule_user(l, hdr);
1074 }
1075
1076 if (pkt_cnt > 1) {
1077 l->stats.sent_fragmented++;
1078 l->stats.sent_fragments += pkt_cnt;
1079 }
1080
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)) {
1084 hdr = buf_msg(skb);
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);
1089 if (!_skb) {
1090 kfree_skb(skb);
1091 __skb_queue_purge(list);
1092 return -ENOBUFS;
1093 }
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;
1098 l->rcv_unacked = 0;
1099 l->stats.sent_pkts++;
1100 seqno++;
1101 continue;
1102 }
1103 if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1104 mss, l->addr, &new_bundle)) {
1105 if (skb) {
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);
1110 } else {
1111 if (new_bundle) {
1112 l->stats.sent_bundles++;
1113 l->stats.sent_bundled++;
1114 }
1115 l->stats.sent_bundled++;
1116 }
1117 continue;
1118 }
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);
1123 }
1124 l->snd_nxt = seqno;
1125 return rc;
1126}
1127
1128static void tipc_link_update_cwin(struct tipc_link *l, int released,
1129 bool retransmitted)
1130{
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;
1135
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);
1140 return;
1141 }
1142 /* Enter slow start */
1143 if (unlikely(!released)) {
1144 l->ssthresh = max_t(u16, l->window / 2, 300);
1145 l->window = l->min_win;
1146 return;
1147 }
1148 /* Don't increase window if no pressure on the transmit queue */
1149 if (txq_len + bklog_len < cwin)
1150 return;
1151
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)
1154 return;
1155
1156 l->cong_acks += released;
1157
1158 /* Slow start */
1159 if (cwin <= l->ssthresh) {
1160 l->window = min_t(u16, cwin + released, l->max_win);
1161 return;
1162 }
1163 /* Congestion avoidance */
1164 if (l->cong_acks < cwin)
1165 return;
1166 l->window = min_t(u16, ++cwin, l->max_win);
1167 l->cong_acks = 0;
1168}
1169
1170static void tipc_link_advance_backlog(struct tipc_link *l,
1171 struct sk_buff_head *xmitq)
1172{
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;
1180 u32 imp;
1181
1182 while (skb_queue_len(txq) < cwin) {
1183 skb = skb_peek(&l->backlogq);
1184 if (!skb)
1185 break;
1186 _skb = skb_clone(skb, GFP_ATOMIC);
1187 if (!_skb)
1188 break;
1189 __skb_dequeue(&l->backlogq);
1190 hdr = buf_msg(skb);
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);
1197
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);
1203 l->rcv_unacked = 0;
1204 l->stats.sent_pkts++;
1205 seqno++;
1206 }
1207 l->snd_nxt = seqno;
1208}
1209
1210/**
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
1215 *
1216 * Return: true if the repeated retransmit failures happens, otherwise
1217 * false
1218 */
1219static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1220 int *rc)
1221{
1222 struct sk_buff *skb = skb_peek(&l->transmq);
1223 struct tipc_msg *hdr;
1224
1225 if (!skb)
1226 return false;
1227
1228 if (!TIPC_SKB_CB(skb)->retr_cnt)
1229 return false;
1230
1231 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1232 msecs_to_jiffies(r->tolerance * 10)))
1233 return false;
1234
1235 hdr = buf_msg(skb);
1236 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1237 return false;
1238
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);
1248
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!");
1252
1253 if (link_is_bc_sndlink(l)) {
1254 r->state = LINK_RESET;
1255 *rc |= TIPC_LINK_DOWN_EVT;
1256 } else {
1257 *rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1258 }
1259
1260 return true;
1261}
1262
1263/* tipc_data_input - deliver data and name distr msgs to upper layer
1264 *
1265 * Consumes buffer if message is of right type
1266 * Node lock must be held
1267 */
1268static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1269 struct sk_buff_head *inputq)
1270{
1271 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1272 struct tipc_msg *hdr = buf_msg(skb);
1273
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);
1281 return true;
1282 }
1283 fallthrough;
1284 case CONN_MANAGER:
1285 skb_queue_tail(inputq, skb);
1286 return true;
1287 case GROUP_PROTOCOL:
1288 skb_queue_tail(mc_inputq, skb);
1289 return true;
1290 case NAME_DISTRIBUTOR:
1291 l->bc_rcvlink->state = LINK_ESTABLISHED;
1292 skb_queue_tail(l->namedq, skb);
1293 return true;
1294 case MSG_BUNDLER:
1295 case TUNNEL_PROTOCOL:
1296 case MSG_FRAGMENTER:
1297 case BCAST_PROTOCOL:
1298 return false;
1299#ifdef CONFIG_TIPC_CRYPTO
1300 case MSG_CRYPTO:
1301 if (sysctl_tipc_key_exchange_enabled &&
1302 TIPC_SKB_CB(skb)->decrypted) {
1303 tipc_crypto_msg_rcv(l->net, skb);
1304 return true;
1305 }
1306 fallthrough;
1307#endif
1308 default:
1309 pr_warn("Dropping received illegal msg type\n");
1310 kfree_skb(skb);
1311 return true;
1312 }
1313}
1314
1315/* tipc_link_input - process packet that has passed link protocol check
1316 *
1317 * Consumes buffer
1318 */
1319static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1320 struct sk_buff_head *inputq,
1321 struct sk_buff **reasm_skb)
1322{
1323 struct tipc_msg *hdr = buf_msg(skb);
1324 struct sk_buff *iskb;
1325 struct sk_buff_head tmpq;
1326 int usr = msg_user(hdr);
1327 int pos = 0;
1328
1329 if (usr == MSG_BUNDLER) {
1330 skb_queue_head_init(&tmpq);
1331 l->stats.recv_bundles++;
1332 l->stats.recv_bundled += msg_msgcnt(hdr);
1333 while (tipc_msg_extract(skb, &iskb, &pos))
1334 tipc_data_input(l, iskb, &tmpq);
1335 tipc_skb_queue_splice_tail(&tmpq, inputq);
1336 return 0;
1337 } else if (usr == MSG_FRAGMENTER) {
1338 l->stats.recv_fragments++;
1339 if (tipc_buf_append(reasm_skb, &skb)) {
1340 l->stats.recv_fragmented++;
1341 tipc_data_input(l, skb, inputq);
1342 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1343 pr_warn_ratelimited("Unable to build fragment list\n");
1344 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1345 }
1346 return 0;
1347 } else if (usr == BCAST_PROTOCOL) {
1348 tipc_bcast_lock(l->net);
1349 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1350 tipc_bcast_unlock(l->net);
1351 }
1352
1353 kfree_skb(skb);
1354 return 0;
1355}
1356
1357/* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1358 * inner message along with the ones in the old link's
1359 * deferdq
1360 * @l: tunnel link
1361 * @skb: TUNNEL_PROTOCOL message
1362 * @inputq: queue to put messages ready for delivery
1363 */
1364static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1365 struct sk_buff_head *inputq)
1366{
1367 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1368 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1369 struct sk_buff_head *fdefq = &l->failover_deferdq;
1370 struct tipc_msg *hdr = buf_msg(skb);
1371 struct sk_buff *iskb;
1372 int ipos = 0;
1373 int rc = 0;
1374 u16 seqno;
1375
1376 if (msg_type(hdr) == SYNCH_MSG) {
1377 kfree_skb(skb);
1378 return 0;
1379 }
1380
1381 /* Not a fragment? */
1382 if (likely(!msg_nof_fragms(hdr))) {
1383 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1384 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1385 skb_queue_len(fdefq));
1386 return 0;
1387 }
1388 kfree_skb(skb);
1389 } else {
1390 /* Set fragment type for buf_append */
1391 if (msg_fragm_no(hdr) == 1)
1392 msg_set_type(hdr, FIRST_FRAGMENT);
1393 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1394 msg_set_type(hdr, FRAGMENT);
1395 else
1396 msg_set_type(hdr, LAST_FRAGMENT);
1397
1398 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1399 /* Successful but non-complete reassembly? */
1400 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1401 return 0;
1402 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1403 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1404 }
1405 iskb = skb;
1406 }
1407
1408 do {
1409 seqno = buf_seqno(iskb);
1410 if (unlikely(less(seqno, l->drop_point))) {
1411 kfree_skb(iskb);
1412 continue;
1413 }
1414 if (unlikely(seqno != l->drop_point)) {
1415 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1416 continue;
1417 }
1418
1419 l->drop_point++;
1420 if (!tipc_data_input(l, iskb, inputq))
1421 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1422 if (unlikely(rc))
1423 break;
1424 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1425
1426 return rc;
1427}
1428
1429/**
1430 * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1431 * @ga: returned pointer to the Gap ACK blocks if any
1432 * @l: the tipc link
1433 * @hdr: the PROTOCOL/STATE_MSG header
1434 * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1435 *
1436 * Return: the total Gap ACK blocks size
1437 */
1438u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1439 struct tipc_msg *hdr, bool uc)
1440{
1441 struct tipc_gap_ack_blks *p;
1442 u16 sz = 0;
1443
1444 /* Does peer support the Gap ACK blocks feature? */
1445 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1446 p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1447 sz = ntohs(p->len);
1448 /* Sanity check */
1449 if (sz == struct_size(p, gacks, p->ugack_cnt + p->bgack_cnt)) {
1450 /* Good, check if the desired type exists */
1451 if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1452 goto ok;
1453 /* Backward compatible: peer might not support bc, but uc? */
1454 } else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1455 if (p->ugack_cnt) {
1456 p->bgack_cnt = 0;
1457 goto ok;
1458 }
1459 }
1460 }
1461 /* Other cases: ignore! */
1462 p = NULL;
1463
1464ok:
1465 *ga = p;
1466 return sz;
1467}
1468
1469static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1470 struct tipc_link *l, u8 start_index)
1471{
1472 struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1473 struct sk_buff *skb = skb_peek(&l->deferdq);
1474 u16 expect, seqno = 0;
1475 u8 n = 0;
1476
1477 if (!skb)
1478 return 0;
1479
1480 expect = buf_seqno(skb);
1481 skb_queue_walk(&l->deferdq, skb) {
1482 seqno = buf_seqno(skb);
1483 if (unlikely(more(seqno, expect))) {
1484 gacks[n].ack = htons(expect - 1);
1485 gacks[n].gap = htons(seqno - expect);
1486 if (++n >= MAX_GAP_ACK_BLKS / 2) {
1487 pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1488 l->name, n,
1489 skb_queue_len(&l->deferdq));
1490 return n;
1491 }
1492 } else if (unlikely(less(seqno, expect))) {
1493 pr_warn("Unexpected skb in deferdq!\n");
1494 continue;
1495 }
1496 expect = seqno + 1;
1497 }
1498
1499 /* last block */
1500 gacks[n].ack = htons(seqno);
1501 gacks[n].gap = 0;
1502 n++;
1503 return n;
1504}
1505
1506/* tipc_build_gap_ack_blks - build Gap ACK blocks
1507 * @l: tipc unicast link
1508 * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1509 *
1510 * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1511 * links of a certain peer, the buffer after built has the network data format
1512 * as found at the struct tipc_gap_ack_blks definition.
1513 *
1514 * returns the actual allocated memory size
1515 */
1516static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1517{
1518 struct tipc_link *bcl = l->bc_rcvlink;
1519 struct tipc_gap_ack_blks *ga;
1520 u16 len;
1521
1522 ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1523
1524 /* Start with broadcast link first */
1525 tipc_bcast_lock(bcl->net);
1526 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1527 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1528 ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1529 tipc_bcast_unlock(bcl->net);
1530
1531 /* Now for unicast link, but an explicit NACK only (???) */
1532 ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1533 __tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1534
1535 /* Total len */
1536 len = struct_size(ga, gacks, ga->bgack_cnt + ga->ugack_cnt);
1537 ga->len = htons(len);
1538 return len;
1539}
1540
1541/* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1542 * acked packets, also doing retransmissions if
1543 * gaps found
1544 * @l: tipc link with transmq queue to be advanced
1545 * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1546 * @acked: seqno of last packet acked by peer without any gaps before
1547 * @gap: # of gap packets
1548 * @ga: buffer pointer to Gap ACK blocks from peer
1549 * @xmitq: queue for accumulating the retransmitted packets if any
1550 * @retransmitted: returned boolean value if a retransmission is really issued
1551 * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1552 * happens (- unlikely case)
1553 *
1554 * Return: the number of packets released from the link transmq
1555 */
1556static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1557 u16 acked, u16 gap,
1558 struct tipc_gap_ack_blks *ga,
1559 struct sk_buff_head *xmitq,
1560 bool *retransmitted, int *rc)
1561{
1562 struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1563 struct tipc_gap_ack *gacks = NULL;
1564 struct sk_buff *skb, *_skb, *tmp;
1565 struct tipc_msg *hdr;
1566 u32 qlen = skb_queue_len(&l->transmq);
1567 u16 nacked = acked, ngap = gap, gack_cnt = 0;
1568 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1569 u16 ack = l->rcv_nxt - 1;
1570 u16 seqno, n = 0;
1571 u16 end = r->acked, start = end, offset = r->last_gap;
1572 u16 si = (last_ga) ? last_ga->start_index : 0;
1573 bool is_uc = !link_is_bc_sndlink(l);
1574 bool bc_has_acked = false;
1575
1576 trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1577
1578 /* Determine Gap ACK blocks if any for the particular link */
1579 if (ga && is_uc) {
1580 /* Get the Gap ACKs, uc part */
1581 gack_cnt = ga->ugack_cnt;
1582 gacks = &ga->gacks[ga->bgack_cnt];
1583 } else if (ga) {
1584 /* Copy the Gap ACKs, bc part, for later renewal if needed */
1585 this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1586 GFP_ATOMIC);
1587 if (likely(this_ga)) {
1588 this_ga->start_index = 0;
1589 /* Start with the bc Gap ACKs */
1590 gack_cnt = this_ga->bgack_cnt;
1591 gacks = &this_ga->gacks[0];
1592 } else {
1593 /* Hmm, we can get in trouble..., simply ignore it */
1594 pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1595 }
1596 }
1597
1598 /* Advance the link transmq */
1599 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1600 seqno = buf_seqno(skb);
1601
1602next_gap_ack:
1603 if (less_eq(seqno, nacked)) {
1604 if (is_uc)
1605 goto release;
1606 /* Skip packets peer has already acked */
1607 if (!more(seqno, r->acked))
1608 continue;
1609 /* Get the next of last Gap ACK blocks */
1610 while (more(seqno, end)) {
1611 if (!last_ga || si >= last_ga->bgack_cnt)
1612 break;
1613 start = end + offset + 1;
1614 end = ntohs(last_ga->gacks[si].ack);
1615 offset = ntohs(last_ga->gacks[si].gap);
1616 si++;
1617 WARN_ONCE(more(start, end) ||
1618 (!offset &&
1619 si < last_ga->bgack_cnt) ||
1620 si > MAX_GAP_ACK_BLKS,
1621 "Corrupted Gap ACK: %d %d %d %d %d\n",
1622 start, end, offset, si,
1623 last_ga->bgack_cnt);
1624 }
1625 /* Check against the last Gap ACK block */
1626 if (in_range(seqno, start, end))
1627 continue;
1628 /* Update/release the packet peer is acking */
1629 bc_has_acked = true;
1630 if (--TIPC_SKB_CB(skb)->ackers)
1631 continue;
1632release:
1633 /* release skb */
1634 __skb_unlink(skb, &l->transmq);
1635 kfree_skb(skb);
1636 } else if (less_eq(seqno, nacked + ngap)) {
1637 /* First gap: check if repeated retrans failures? */
1638 if (unlikely(seqno == acked + 1 &&
1639 link_retransmit_failure(l, r, rc))) {
1640 /* Ignore this bc Gap ACKs if any */
1641 kfree(this_ga);
1642 this_ga = NULL;
1643 break;
1644 }
1645 /* retransmit skb if unrestricted*/
1646 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1647 continue;
1648 tipc_link_set_skb_retransmit_time(skb, l);
1649 _skb = pskb_copy(skb, GFP_ATOMIC);
1650 if (!_skb)
1651 continue;
1652 hdr = buf_msg(_skb);
1653 msg_set_ack(hdr, ack);
1654 msg_set_bcast_ack(hdr, bc_ack);
1655 _skb->priority = TC_PRIO_CONTROL;
1656 __skb_queue_tail(xmitq, _skb);
1657 l->stats.retransmitted++;
1658 if (!is_uc)
1659 r->stats.retransmitted++;
1660 *retransmitted = true;
1661 /* Increase actual retrans counter & mark first time */
1662 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1663 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1664 } else {
1665 /* retry with Gap ACK blocks if any */
1666 if (n >= gack_cnt)
1667 break;
1668 nacked = ntohs(gacks[n].ack);
1669 ngap = ntohs(gacks[n].gap);
1670 n++;
1671 goto next_gap_ack;
1672 }
1673 }
1674
1675 /* Renew last Gap ACK blocks for bc if needed */
1676 if (bc_has_acked) {
1677 if (this_ga) {
1678 kfree(last_ga);
1679 r->last_ga = this_ga;
1680 r->last_gap = gap;
1681 } else if (last_ga) {
1682 if (less(acked, start)) {
1683 si--;
1684 offset = start - acked - 1;
1685 } else if (less(acked, end)) {
1686 acked = end;
1687 }
1688 if (si < last_ga->bgack_cnt) {
1689 last_ga->start_index = si;
1690 r->last_gap = offset;
1691 } else {
1692 kfree(last_ga);
1693 r->last_ga = NULL;
1694 r->last_gap = 0;
1695 }
1696 } else {
1697 r->last_gap = 0;
1698 }
1699 r->acked = acked;
1700 } else {
1701 kfree(this_ga);
1702 }
1703
1704 return qlen - skb_queue_len(&l->transmq);
1705}
1706
1707/* tipc_link_build_state_msg: prepare link state message for transmission
1708 *
1709 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1710 * risk of ack storms towards the sender
1711 */
1712int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1713{
1714 if (!l)
1715 return 0;
1716
1717 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1718 if (link_is_bc_rcvlink(l)) {
1719 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1720 return 0;
1721 l->rcv_unacked = 0;
1722
1723 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1724 l->snd_nxt = l->rcv_nxt;
1725 return TIPC_LINK_SND_STATE;
1726 }
1727 /* Unicast ACK */
1728 l->rcv_unacked = 0;
1729 l->stats.sent_acks++;
1730 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1731 return 0;
1732}
1733
1734/* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1735 */
1736void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1737{
1738 int mtyp = RESET_MSG;
1739 struct sk_buff *skb;
1740
1741 if (l->state == LINK_ESTABLISHING)
1742 mtyp = ACTIVATE_MSG;
1743
1744 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1745
1746 /* Inform peer that this endpoint is going down if applicable */
1747 skb = skb_peek_tail(xmitq);
1748 if (skb && (l->state == LINK_RESET))
1749 msg_set_peer_stopping(buf_msg(skb), 1);
1750}
1751
1752/* tipc_link_build_nack_msg: prepare link nack message for transmission
1753 * Note that sending of broadcast NACK is coordinated among nodes, to
1754 * reduce the risk of NACK storms towards the sender
1755 */
1756static int tipc_link_build_nack_msg(struct tipc_link *l,
1757 struct sk_buff_head *xmitq)
1758{
1759 u32 def_cnt = ++l->stats.deferred_recv;
1760 struct sk_buff_head *dfq = &l->deferdq;
1761 u32 defq_len = skb_queue_len(dfq);
1762 int match1, match2;
1763
1764 if (link_is_bc_rcvlink(l)) {
1765 match1 = def_cnt & 0xf;
1766 match2 = tipc_own_addr(l->net) & 0xf;
1767 if (match1 == match2)
1768 return TIPC_LINK_SND_STATE;
1769 return 0;
1770 }
1771
1772 if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1773 u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1774
1775 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1776 rcvgap, 0, 0, xmitq);
1777 }
1778 return 0;
1779}
1780
1781/* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1782 * @l: the link that should handle the message
1783 * @skb: TIPC packet
1784 * @xmitq: queue to place packets to be sent after this call
1785 */
1786int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1787 struct sk_buff_head *xmitq)
1788{
1789 struct sk_buff_head *defq = &l->deferdq;
1790 struct tipc_msg *hdr = buf_msg(skb);
1791 u16 seqno, rcv_nxt, win_lim;
1792 int released = 0;
1793 int rc = 0;
1794
1795 /* Verify and update link state */
1796 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1797 return tipc_link_proto_rcv(l, skb, xmitq);
1798
1799 /* Don't send probe at next timeout expiration */
1800 l->silent_intv_cnt = 0;
1801
1802 do {
1803 hdr = buf_msg(skb);
1804 seqno = msg_seqno(hdr);
1805 rcv_nxt = l->rcv_nxt;
1806 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1807
1808 if (unlikely(!link_is_up(l))) {
1809 if (l->state == LINK_ESTABLISHING)
1810 rc = TIPC_LINK_UP_EVT;
1811 kfree_skb(skb);
1812 break;
1813 }
1814
1815 /* Drop if outside receive window */
1816 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1817 l->stats.duplicates++;
1818 kfree_skb(skb);
1819 break;
1820 }
1821 released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1822 NULL, NULL, NULL, NULL);
1823
1824 /* Defer delivery if sequence gap */
1825 if (unlikely(seqno != rcv_nxt)) {
1826 if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1827 l->stats.duplicates++;
1828 rc |= tipc_link_build_nack_msg(l, xmitq);
1829 break;
1830 }
1831
1832 /* Deliver packet */
1833 l->rcv_nxt++;
1834 l->stats.recv_pkts++;
1835
1836 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1837 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1838 else if (!tipc_data_input(l, skb, l->inputq))
1839 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1840 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1841 rc |= tipc_link_build_state_msg(l, xmitq);
1842 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1843 break;
1844 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1845
1846 /* Forward queues and wake up waiting users */
1847 if (released) {
1848 tipc_link_update_cwin(l, released, 0);
1849 tipc_link_advance_backlog(l, xmitq);
1850 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1851 link_prepare_wakeup(l);
1852 }
1853 return rc;
1854}
1855
1856static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1857 bool probe_reply, u16 rcvgap,
1858 int tolerance, int priority,
1859 struct sk_buff_head *xmitq)
1860{
1861 struct tipc_mon_state *mstate = &l->mon_state;
1862 struct sk_buff_head *dfq = &l->deferdq;
1863 struct tipc_link *bcl = l->bc_rcvlink;
1864 struct tipc_msg *hdr;
1865 struct sk_buff *skb;
1866 bool node_up = link_is_up(bcl);
1867 u16 glen = 0, bc_rcvgap = 0;
1868 int dlen = 0;
1869 void *data;
1870
1871 /* Don't send protocol message during reset or link failover */
1872 if (tipc_link_is_blocked(l))
1873 return;
1874
1875 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1876 return;
1877
1878 if ((probe || probe_reply) && !skb_queue_empty(dfq))
1879 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1880
1881 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1882 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1883 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1884 if (!skb)
1885 return;
1886
1887 hdr = buf_msg(skb);
1888 data = msg_data(hdr);
1889 msg_set_session(hdr, l->session);
1890 msg_set_bearer_id(hdr, l->bearer_id);
1891 msg_set_net_plane(hdr, l->net_plane);
1892 msg_set_next_sent(hdr, l->snd_nxt);
1893 msg_set_ack(hdr, l->rcv_nxt - 1);
1894 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1895 msg_set_bc_ack_invalid(hdr, !node_up);
1896 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1897 msg_set_link_tolerance(hdr, tolerance);
1898 msg_set_linkprio(hdr, priority);
1899 msg_set_redundant_link(hdr, node_up);
1900 msg_set_seq_gap(hdr, 0);
1901 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1902
1903 if (mtyp == STATE_MSG) {
1904 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1905 msg_set_seqno(hdr, l->snd_nxt_state++);
1906 msg_set_seq_gap(hdr, rcvgap);
1907 bc_rcvgap = link_bc_rcv_gap(bcl);
1908 msg_set_bc_gap(hdr, bc_rcvgap);
1909 msg_set_probe(hdr, probe);
1910 msg_set_is_keepalive(hdr, probe || probe_reply);
1911 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1912 glen = tipc_build_gap_ack_blks(l, hdr);
1913 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1914 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1915 skb_trim(skb, INT_H_SIZE + glen + dlen);
1916 l->stats.sent_states++;
1917 l->rcv_unacked = 0;
1918 } else {
1919 /* RESET_MSG or ACTIVATE_MSG */
1920 if (mtyp == ACTIVATE_MSG) {
1921 msg_set_dest_session_valid(hdr, 1);
1922 msg_set_dest_session(hdr, l->peer_session);
1923 }
1924 msg_set_max_pkt(hdr, l->advertised_mtu);
1925 strcpy(data, l->if_name);
1926 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1927 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1928 }
1929 if (probe)
1930 l->stats.sent_probes++;
1931 if (rcvgap)
1932 l->stats.sent_nacks++;
1933 if (bc_rcvgap)
1934 bcl->stats.sent_nacks++;
1935 skb->priority = TC_PRIO_CONTROL;
1936 __skb_queue_tail(xmitq, skb);
1937 trace_tipc_proto_build(skb, false, l->name);
1938}
1939
1940void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1941 struct sk_buff_head *xmitq)
1942{
1943 u32 onode = tipc_own_addr(l->net);
1944 struct tipc_msg *hdr, *ihdr;
1945 struct sk_buff_head tnlq;
1946 struct sk_buff *skb;
1947 u32 dnode = l->addr;
1948
1949 __skb_queue_head_init(&tnlq);
1950 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1951 INT_H_SIZE, BASIC_H_SIZE,
1952 dnode, onode, 0, 0, 0);
1953 if (!skb) {
1954 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1955 return;
1956 }
1957
1958 hdr = buf_msg(skb);
1959 msg_set_msgcnt(hdr, 1);
1960 msg_set_bearer_id(hdr, l->peer_bearer_id);
1961
1962 ihdr = (struct tipc_msg *)msg_data(hdr);
1963 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1964 BASIC_H_SIZE, dnode);
1965 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1966 __skb_queue_tail(&tnlq, skb);
1967 tipc_link_xmit(l, &tnlq, xmitq);
1968}
1969
1970/* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1971 * with contents of the link's transmit and backlog queues.
1972 */
1973void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1974 int mtyp, struct sk_buff_head *xmitq)
1975{
1976 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1977 struct sk_buff *skb, *tnlskb;
1978 struct tipc_msg *hdr, tnlhdr;
1979 struct sk_buff_head *queue = &l->transmq;
1980 struct sk_buff_head tmpxq, tnlq, frags;
1981 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1982 bool pktcnt_need_update = false;
1983 u16 syncpt;
1984 int rc;
1985
1986 if (!tnl)
1987 return;
1988
1989 __skb_queue_head_init(&tnlq);
1990 /* Link Synching:
1991 * From now on, send only one single ("dummy") SYNCH message
1992 * to peer. The SYNCH message does not contain any data, just
1993 * a header conveying the synch point to the peer.
1994 */
1995 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1996 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1997 INT_H_SIZE, 0, l->addr,
1998 tipc_own_addr(l->net),
1999 0, 0, 0);
2000 if (!tnlskb) {
2001 pr_warn("%sunable to create dummy SYNCH_MSG\n",
2002 link_co_err);
2003 return;
2004 }
2005
2006 hdr = buf_msg(tnlskb);
2007 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
2008 msg_set_syncpt(hdr, syncpt);
2009 msg_set_bearer_id(hdr, l->peer_bearer_id);
2010 __skb_queue_tail(&tnlq, tnlskb);
2011 tipc_link_xmit(tnl, &tnlq, xmitq);
2012 return;
2013 }
2014
2015 __skb_queue_head_init(&tmpxq);
2016 __skb_queue_head_init(&frags);
2017 /* At least one packet required for safe algorithm => add dummy */
2018 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
2019 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
2020 0, 0, TIPC_ERR_NO_PORT);
2021 if (!skb) {
2022 pr_warn("%sunable to create tunnel packet\n", link_co_err);
2023 return;
2024 }
2025 __skb_queue_tail(&tnlq, skb);
2026 tipc_link_xmit(l, &tnlq, &tmpxq);
2027 __skb_queue_purge(&tmpxq);
2028
2029 /* Initialize reusable tunnel packet header */
2030 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
2031 mtyp, INT_H_SIZE, l->addr);
2032 if (mtyp == SYNCH_MSG)
2033 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
2034 else
2035 pktcnt = skb_queue_len(&l->transmq);
2036 pktcnt += skb_queue_len(&l->backlogq);
2037 msg_set_msgcnt(&tnlhdr, pktcnt);
2038 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
2039tnl:
2040 /* Wrap each packet into a tunnel packet */
2041 skb_queue_walk(queue, skb) {
2042 hdr = buf_msg(skb);
2043 if (queue == &l->backlogq)
2044 msg_set_seqno(hdr, seqno++);
2045 pktlen = msg_size(hdr);
2046
2047 /* Tunnel link MTU is not large enough? This could be
2048 * due to:
2049 * 1) Link MTU has just changed or set differently;
2050 * 2) Or FAILOVER on the top of a SYNCH message
2051 *
2052 * The 2nd case should not happen if peer supports
2053 * TIPC_TUNNEL_ENHANCED
2054 */
2055 if (pktlen > tnl->mtu - INT_H_SIZE) {
2056 if (mtyp == FAILOVER_MSG &&
2057 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2058 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2059 &frags);
2060 if (rc) {
2061 pr_warn("%sunable to frag msg: rc %d\n",
2062 link_co_err, rc);
2063 return;
2064 }
2065 pktcnt += skb_queue_len(&frags) - 1;
2066 pktcnt_need_update = true;
2067 skb_queue_splice_tail_init(&frags, &tnlq);
2068 continue;
2069 }
2070 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2071 * => Just warn it and return!
2072 */
2073 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2074 link_co_err, msg_user(hdr),
2075 msg_type(hdr), msg_size(hdr));
2076 return;
2077 }
2078
2079 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2080 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2081 if (!tnlskb) {
2082 pr_warn("%sunable to send packet\n", link_co_err);
2083 return;
2084 }
2085 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2086 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2087 __skb_queue_tail(&tnlq, tnlskb);
2088 }
2089 if (queue != &l->backlogq) {
2090 queue = &l->backlogq;
2091 goto tnl;
2092 }
2093
2094 if (pktcnt_need_update)
2095 skb_queue_walk(&tnlq, skb) {
2096 hdr = buf_msg(skb);
2097 msg_set_msgcnt(hdr, pktcnt);
2098 }
2099
2100 tipc_link_xmit(tnl, &tnlq, xmitq);
2101
2102 if (mtyp == FAILOVER_MSG) {
2103 tnl->drop_point = l->rcv_nxt;
2104 tnl->failover_reasm_skb = l->reasm_buf;
2105 l->reasm_buf = NULL;
2106
2107 /* Failover the link's deferdq */
2108 if (unlikely(!skb_queue_empty(fdefq))) {
2109 pr_warn("Link failover deferdq not empty: %d!\n",
2110 skb_queue_len(fdefq));
2111 __skb_queue_purge(fdefq);
2112 }
2113 skb_queue_splice_init(&l->deferdq, fdefq);
2114 }
2115}
2116
2117/**
2118 * tipc_link_failover_prepare() - prepare tnl for link failover
2119 *
2120 * This is a special version of the precursor - tipc_link_tnl_prepare(),
2121 * see the tipc_node_link_failover() for details
2122 *
2123 * @l: failover link
2124 * @tnl: tunnel link
2125 * @xmitq: queue for messages to be xmited
2126 */
2127void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2128 struct sk_buff_head *xmitq)
2129{
2130 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2131
2132 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2133
2134 /* This failover link endpoint was never established before,
2135 * so it has not received anything from peer.
2136 * Otherwise, it must be a normal failover situation or the
2137 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2138 * would have to start over from scratch instead.
2139 */
2140 tnl->drop_point = 1;
2141 tnl->failover_reasm_skb = NULL;
2142
2143 /* Initiate the link's failover deferdq */
2144 if (unlikely(!skb_queue_empty(fdefq))) {
2145 pr_warn("Link failover deferdq not empty: %d!\n",
2146 skb_queue_len(fdefq));
2147 __skb_queue_purge(fdefq);
2148 }
2149}
2150
2151/* tipc_link_validate_msg(): validate message against current link state
2152 * Returns true if message should be accepted, otherwise false
2153 */
2154bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2155{
2156 u16 curr_session = l->peer_session;
2157 u16 session = msg_session(hdr);
2158 int mtyp = msg_type(hdr);
2159
2160 if (msg_user(hdr) != LINK_PROTOCOL)
2161 return true;
2162
2163 switch (mtyp) {
2164 case RESET_MSG:
2165 if (!l->in_session)
2166 return true;
2167 /* Accept only RESET with new session number */
2168 return more(session, curr_session);
2169 case ACTIVATE_MSG:
2170 if (!l->in_session)
2171 return true;
2172 /* Accept only ACTIVATE with new or current session number */
2173 return !less(session, curr_session);
2174 case STATE_MSG:
2175 /* Accept only STATE with current session number */
2176 if (!l->in_session)
2177 return false;
2178 if (session != curr_session)
2179 return false;
2180 /* Extra sanity check */
2181 if (!link_is_up(l) && msg_ack(hdr))
2182 return false;
2183 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2184 return true;
2185 /* Accept only STATE with new sequence number */
2186 return !less(msg_seqno(hdr), l->rcv_nxt_state);
2187 default:
2188 return false;
2189 }
2190}
2191
2192/* tipc_link_proto_rcv(): receive link level protocol message :
2193 * Note that network plane id propagates through the network, and may
2194 * change at any time. The node with lowest numerical id determines
2195 * network plane
2196 */
2197static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2198 struct sk_buff_head *xmitq)
2199{
2200 struct tipc_msg *hdr = buf_msg(skb);
2201 struct tipc_gap_ack_blks *ga = NULL;
2202 bool reply = msg_probe(hdr), retransmitted = false;
2203 u32 dlen = msg_data_sz(hdr), glen = 0;
2204 u16 peers_snd_nxt = msg_next_sent(hdr);
2205 u16 peers_tol = msg_link_tolerance(hdr);
2206 u16 peers_prio = msg_linkprio(hdr);
2207 u16 gap = msg_seq_gap(hdr);
2208 u16 ack = msg_ack(hdr);
2209 u16 rcv_nxt = l->rcv_nxt;
2210 u16 rcvgap = 0;
2211 int mtyp = msg_type(hdr);
2212 int rc = 0, released;
2213 char *if_name;
2214 void *data;
2215
2216 trace_tipc_proto_rcv(skb, false, l->name);
2217
2218 if (dlen > U16_MAX)
2219 goto exit;
2220
2221 if (tipc_link_is_blocked(l) || !xmitq)
2222 goto exit;
2223
2224 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2225 l->net_plane = msg_net_plane(hdr);
2226
2227 if (skb_linearize(skb))
2228 goto exit;
2229
2230 hdr = buf_msg(skb);
2231 data = msg_data(hdr);
2232
2233 if (!tipc_link_validate_msg(l, hdr)) {
2234 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2235 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2236 goto exit;
2237 }
2238
2239 switch (mtyp) {
2240 case RESET_MSG:
2241 case ACTIVATE_MSG:
2242 /* Complete own link name with peer's interface name */
2243 if_name = strrchr(l->name, ':') + 1;
2244 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2245 break;
2246 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2247 break;
2248 strncpy(if_name, data, TIPC_MAX_IF_NAME);
2249
2250 /* Update own tolerance if peer indicates a non-zero value */
2251 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2252 l->tolerance = peers_tol;
2253 l->bc_rcvlink->tolerance = peers_tol;
2254 }
2255 /* Update own priority if peer's priority is higher */
2256 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2257 l->priority = peers_prio;
2258
2259 /* If peer is going down we want full re-establish cycle */
2260 if (msg_peer_stopping(hdr)) {
2261 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2262 break;
2263 }
2264
2265 /* If this endpoint was re-created while peer was ESTABLISHING
2266 * it doesn't know current session number. Force re-synch.
2267 */
2268 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2269 l->session != msg_dest_session(hdr)) {
2270 if (less(l->session, msg_dest_session(hdr)))
2271 l->session = msg_dest_session(hdr) + 1;
2272 break;
2273 }
2274
2275 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2276 if (mtyp == RESET_MSG || !link_is_up(l))
2277 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2278
2279 /* ACTIVATE_MSG takes up link if it was already locally reset */
2280 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2281 rc = TIPC_LINK_UP_EVT;
2282
2283 l->peer_session = msg_session(hdr);
2284 l->in_session = true;
2285 l->peer_bearer_id = msg_bearer_id(hdr);
2286 if (l->mtu > msg_max_pkt(hdr))
2287 l->mtu = msg_max_pkt(hdr);
2288 break;
2289
2290 case STATE_MSG:
2291 /* Validate Gap ACK blocks, drop if invalid */
2292 glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2293 if (glen > dlen)
2294 break;
2295
2296 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2297
2298 /* Update own tolerance if peer indicates a non-zero value */
2299 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2300 l->tolerance = peers_tol;
2301 l->bc_rcvlink->tolerance = peers_tol;
2302 }
2303 /* Update own prio if peer indicates a different value */
2304 if ((peers_prio != l->priority) &&
2305 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2306 l->priority = peers_prio;
2307 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2308 }
2309
2310 l->silent_intv_cnt = 0;
2311 l->stats.recv_states++;
2312 if (msg_probe(hdr))
2313 l->stats.recv_probes++;
2314
2315 if (!link_is_up(l)) {
2316 if (l->state == LINK_ESTABLISHING)
2317 rc = TIPC_LINK_UP_EVT;
2318 break;
2319 }
2320
2321 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2322 &l->mon_state, l->bearer_id);
2323
2324 /* Send NACK if peer has sent pkts we haven't received yet */
2325 if ((reply || msg_is_keepalive(hdr)) &&
2326 more(peers_snd_nxt, rcv_nxt) &&
2327 !tipc_link_is_synching(l) &&
2328 skb_queue_empty(&l->deferdq))
2329 rcvgap = peers_snd_nxt - l->rcv_nxt;
2330 if (rcvgap || reply)
2331 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2332 rcvgap, 0, 0, xmitq);
2333
2334 released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2335 &retransmitted, &rc);
2336 if (gap)
2337 l->stats.recv_nacks++;
2338 if (released || retransmitted)
2339 tipc_link_update_cwin(l, released, retransmitted);
2340 if (released)
2341 tipc_link_advance_backlog(l, xmitq);
2342 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2343 link_prepare_wakeup(l);
2344 }
2345exit:
2346 kfree_skb(skb);
2347 return rc;
2348}
2349
2350/* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2351 */
2352static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2353 u16 peers_snd_nxt,
2354 struct sk_buff_head *xmitq)
2355{
2356 struct sk_buff *skb;
2357 struct tipc_msg *hdr;
2358 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2359 u16 ack = l->rcv_nxt - 1;
2360 u16 gap_to = peers_snd_nxt - 1;
2361
2362 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2363 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2364 if (!skb)
2365 return false;
2366 hdr = buf_msg(skb);
2367 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2368 msg_set_bcast_ack(hdr, ack);
2369 msg_set_bcgap_after(hdr, ack);
2370 if (dfrd_skb)
2371 gap_to = buf_seqno(dfrd_skb) - 1;
2372 msg_set_bcgap_to(hdr, gap_to);
2373 msg_set_non_seq(hdr, bcast);
2374 __skb_queue_tail(xmitq, skb);
2375 return true;
2376}
2377
2378/* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2379 *
2380 * Give a newly added peer node the sequence number where it should
2381 * start receiving and acking broadcast packets.
2382 */
2383static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2384 struct sk_buff_head *xmitq)
2385{
2386 struct sk_buff_head list;
2387
2388 __skb_queue_head_init(&list);
2389 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2390 return;
2391 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2392 tipc_link_xmit(l, &list, xmitq);
2393}
2394
2395/* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2396 */
2397void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2398{
2399 int mtyp = msg_type(hdr);
2400 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2401
2402 if (link_is_up(l))
2403 return;
2404
2405 if (msg_user(hdr) == BCAST_PROTOCOL) {
2406 l->rcv_nxt = peers_snd_nxt;
2407 l->state = LINK_ESTABLISHED;
2408 return;
2409 }
2410
2411 if (l->peer_caps & TIPC_BCAST_SYNCH)
2412 return;
2413
2414 if (msg_peer_node_is_up(hdr))
2415 return;
2416
2417 /* Compatibility: accept older, less safe initial synch data */
2418 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2419 l->rcv_nxt = peers_snd_nxt;
2420}
2421
2422/* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2423 */
2424int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2425 struct sk_buff_head *xmitq)
2426{
2427 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2428 int rc = 0;
2429
2430 if (!link_is_up(l))
2431 return rc;
2432
2433 if (!msg_peer_node_is_up(hdr))
2434 return rc;
2435
2436 /* Open when peer acknowledges our bcast init msg (pkt #1) */
2437 if (msg_ack(hdr))
2438 l->bc_peer_is_up = true;
2439
2440 if (!l->bc_peer_is_up)
2441 return rc;
2442
2443 /* Ignore if peers_snd_nxt goes beyond receive window */
2444 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2445 return rc;
2446
2447 l->snd_nxt = peers_snd_nxt;
2448 if (link_bc_rcv_gap(l))
2449 rc |= TIPC_LINK_SND_STATE;
2450
2451 /* Return now if sender supports nack via STATE messages */
2452 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2453 return rc;
2454
2455 /* Otherwise, be backwards compatible */
2456
2457 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2458 l->nack_state = BC_NACK_SND_CONDITIONAL;
2459 return 0;
2460 }
2461
2462 /* Don't NACK if one was recently sent or peeked */
2463 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2464 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2465 return 0;
2466 }
2467
2468 /* Conditionally delay NACK sending until next synch rcv */
2469 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2470 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2471 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2472 return 0;
2473 }
2474
2475 /* Send NACK now but suppress next one */
2476 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2477 l->nack_state = BC_NACK_SND_SUPPRESS;
2478 return 0;
2479}
2480
2481int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2482 struct tipc_gap_ack_blks *ga,
2483 struct sk_buff_head *xmitq,
2484 struct sk_buff_head *retrq)
2485{
2486 struct tipc_link *l = r->bc_sndlink;
2487 bool unused = false;
2488 int rc = 0;
2489
2490 if (!link_is_up(r) || !r->bc_peer_is_up)
2491 return 0;
2492
2493 if (gap) {
2494 l->stats.recv_nacks++;
2495 r->stats.recv_nacks++;
2496 }
2497
2498 if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2499 return 0;
2500
2501 trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2502 tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2503
2504 tipc_link_advance_backlog(l, xmitq);
2505 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2506 link_prepare_wakeup(l);
2507
2508 return rc;
2509}
2510
2511/* tipc_link_bc_nack_rcv(): receive broadcast nack message
2512 * This function is here for backwards compatibility, since
2513 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2514 */
2515int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2516 struct sk_buff_head *xmitq)
2517{
2518 struct tipc_msg *hdr = buf_msg(skb);
2519 u32 dnode = msg_destnode(hdr);
2520 int mtyp = msg_type(hdr);
2521 u16 acked = msg_bcast_ack(hdr);
2522 u16 from = acked + 1;
2523 u16 to = msg_bcgap_to(hdr);
2524 u16 peers_snd_nxt = to + 1;
2525 int rc = 0;
2526
2527 kfree_skb(skb);
2528
2529 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2530 return 0;
2531
2532 if (mtyp != STATE_MSG)
2533 return 0;
2534
2535 if (dnode == tipc_own_addr(l->net)) {
2536 rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2537 xmitq);
2538 l->stats.recv_nacks++;
2539 return rc;
2540 }
2541
2542 /* Msg for other node => suppress own NACK at next sync if applicable */
2543 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2544 l->nack_state = BC_NACK_SND_SUPPRESS;
2545
2546 return 0;
2547}
2548
2549void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2550{
2551 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2552
2553 l->min_win = min_win;
2554 l->ssthresh = max_win;
2555 l->max_win = max_win;
2556 l->window = min_win;
2557 l->backlog[TIPC_LOW_IMPORTANCE].limit = min_win * 2;
2558 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = min_win * 4;
2559 l->backlog[TIPC_HIGH_IMPORTANCE].limit = min_win * 6;
2560 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2561 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2562}
2563
2564/**
2565 * tipc_link_reset_stats - reset link statistics
2566 * @l: pointer to link
2567 */
2568void tipc_link_reset_stats(struct tipc_link *l)
2569{
2570 memset(&l->stats, 0, sizeof(l->stats));
2571}
2572
2573static void link_print(struct tipc_link *l, const char *str)
2574{
2575 struct sk_buff *hskb = skb_peek(&l->transmq);
2576 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2577 u16 tail = l->snd_nxt - 1;
2578
2579 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2580 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2581 skb_queue_len(&l->transmq), head, tail,
2582 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2583}
2584
2585/* Parse and validate nested (link) properties valid for media, bearer and link
2586 */
2587int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2588{
2589 int err;
2590
2591 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2592 tipc_nl_prop_policy, NULL);
2593 if (err)
2594 return err;
2595
2596 if (props[TIPC_NLA_PROP_PRIO]) {
2597 u32 prio;
2598
2599 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2600 if (prio > TIPC_MAX_LINK_PRI)
2601 return -EINVAL;
2602 }
2603
2604 if (props[TIPC_NLA_PROP_TOL]) {
2605 u32 tol;
2606
2607 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2608 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2609 return -EINVAL;
2610 }
2611
2612 if (props[TIPC_NLA_PROP_WIN]) {
2613 u32 max_win;
2614
2615 max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2616 if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2617 return -EINVAL;
2618 }
2619
2620 return 0;
2621}
2622
2623static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2624{
2625 int i;
2626 struct nlattr *stats;
2627
2628 struct nla_map {
2629 u32 key;
2630 u32 val;
2631 };
2632
2633 struct nla_map map[] = {
2634 {TIPC_NLA_STATS_RX_INFO, 0},
2635 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2636 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2637 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2638 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2639 {TIPC_NLA_STATS_TX_INFO, 0},
2640 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2641 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2642 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2643 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2644 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2645 s->msg_length_counts : 1},
2646 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2647 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2648 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2649 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2650 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2651 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2652 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2653 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2654 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2655 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2656 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2657 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2658 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2659 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2660 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2661 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2662 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2663 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2664 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2665 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2666 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2667 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2668 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2669 };
2670
2671 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2672 if (!stats)
2673 return -EMSGSIZE;
2674
2675 for (i = 0; i < ARRAY_SIZE(map); i++)
2676 if (nla_put_u32(skb, map[i].key, map[i].val))
2677 goto msg_full;
2678
2679 nla_nest_end(skb, stats);
2680
2681 return 0;
2682msg_full:
2683 nla_nest_cancel(skb, stats);
2684
2685 return -EMSGSIZE;
2686}
2687
2688/* Caller should hold appropriate locks to protect the link */
2689int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2690 struct tipc_link *link, int nlflags)
2691{
2692 u32 self = tipc_own_addr(net);
2693 struct nlattr *attrs;
2694 struct nlattr *prop;
2695 void *hdr;
2696 int err;
2697
2698 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2699 nlflags, TIPC_NL_LINK_GET);
2700 if (!hdr)
2701 return -EMSGSIZE;
2702
2703 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2704 if (!attrs)
2705 goto msg_full;
2706
2707 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2708 goto attr_msg_full;
2709 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2710 goto attr_msg_full;
2711 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2712 goto attr_msg_full;
2713 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2714 goto attr_msg_full;
2715 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2716 goto attr_msg_full;
2717
2718 if (tipc_link_is_up(link))
2719 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2720 goto attr_msg_full;
2721 if (link->active)
2722 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2723 goto attr_msg_full;
2724
2725 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2726 if (!prop)
2727 goto attr_msg_full;
2728 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2729 goto prop_msg_full;
2730 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2731 goto prop_msg_full;
2732 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2733 link->window))
2734 goto prop_msg_full;
2735 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2736 goto prop_msg_full;
2737 nla_nest_end(msg->skb, prop);
2738
2739 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2740 if (err)
2741 goto attr_msg_full;
2742
2743 nla_nest_end(msg->skb, attrs);
2744 genlmsg_end(msg->skb, hdr);
2745
2746 return 0;
2747
2748prop_msg_full:
2749 nla_nest_cancel(msg->skb, prop);
2750attr_msg_full:
2751 nla_nest_cancel(msg->skb, attrs);
2752msg_full:
2753 genlmsg_cancel(msg->skb, hdr);
2754
2755 return -EMSGSIZE;
2756}
2757
2758static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2759 struct tipc_stats *stats)
2760{
2761 int i;
2762 struct nlattr *nest;
2763
2764 struct nla_map {
2765 __u32 key;
2766 __u32 val;
2767 };
2768
2769 struct nla_map map[] = {
2770 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2771 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2772 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2773 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2774 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2775 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2776 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2777 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2778 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2779 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2780 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2781 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2782 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2783 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2784 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2785 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2786 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2787 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2788 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2789 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2790 };
2791
2792 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2793 if (!nest)
2794 return -EMSGSIZE;
2795
2796 for (i = 0; i < ARRAY_SIZE(map); i++)
2797 if (nla_put_u32(skb, map[i].key, map[i].val))
2798 goto msg_full;
2799
2800 nla_nest_end(skb, nest);
2801
2802 return 0;
2803msg_full:
2804 nla_nest_cancel(skb, nest);
2805
2806 return -EMSGSIZE;
2807}
2808
2809int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2810 struct tipc_link *bcl)
2811{
2812 int err;
2813 void *hdr;
2814 struct nlattr *attrs;
2815 struct nlattr *prop;
2816 u32 bc_mode = tipc_bcast_get_mode(net);
2817 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2818
2819 if (!bcl)
2820 return 0;
2821
2822 tipc_bcast_lock(net);
2823
2824 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2825 NLM_F_MULTI, TIPC_NL_LINK_GET);
2826 if (!hdr) {
2827 tipc_bcast_unlock(net);
2828 return -EMSGSIZE;
2829 }
2830
2831 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2832 if (!attrs)
2833 goto msg_full;
2834
2835 /* The broadcast link is always up */
2836 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2837 goto attr_msg_full;
2838
2839 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2840 goto attr_msg_full;
2841 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2842 goto attr_msg_full;
2843 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2844 goto attr_msg_full;
2845 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2846 goto attr_msg_full;
2847
2848 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2849 if (!prop)
2850 goto attr_msg_full;
2851 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2852 goto prop_msg_full;
2853 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2854 goto prop_msg_full;
2855 if (bc_mode & BCLINK_MODE_SEL)
2856 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2857 bc_ratio))
2858 goto prop_msg_full;
2859 nla_nest_end(msg->skb, prop);
2860
2861 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2862 if (err)
2863 goto attr_msg_full;
2864
2865 tipc_bcast_unlock(net);
2866 nla_nest_end(msg->skb, attrs);
2867 genlmsg_end(msg->skb, hdr);
2868
2869 return 0;
2870
2871prop_msg_full:
2872 nla_nest_cancel(msg->skb, prop);
2873attr_msg_full:
2874 nla_nest_cancel(msg->skb, attrs);
2875msg_full:
2876 tipc_bcast_unlock(net);
2877 genlmsg_cancel(msg->skb, hdr);
2878
2879 return -EMSGSIZE;
2880}
2881
2882void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2883 struct sk_buff_head *xmitq)
2884{
2885 l->tolerance = tol;
2886 if (l->bc_rcvlink)
2887 l->bc_rcvlink->tolerance = tol;
2888 if (link_is_up(l))
2889 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2890}
2891
2892void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2893 struct sk_buff_head *xmitq)
2894{
2895 l->priority = prio;
2896 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2897}
2898
2899void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2900{
2901 l->abort_limit = limit;
2902}
2903
2904/**
2905 * tipc_link_dump - dump TIPC link data
2906 * @l: tipc link to be dumped
2907 * @dqueues: bitmask to decide if any link queue to be dumped?
2908 * - TIPC_DUMP_NONE: don't dump link queues
2909 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2910 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2911 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2912 * - TIPC_DUMP_INPUTQ: dump link input queue
2913 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2914 * - TIPC_DUMP_ALL: dump all the link queues above
2915 * @buf: returned buffer of dump data in format
2916 */
2917int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2918{
2919 int i = 0;
2920 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2921 struct sk_buff_head *list;
2922 struct sk_buff *hskb, *tskb;
2923 u32 len;
2924
2925 if (!l) {
2926 i += scnprintf(buf, sz, "link data: (null)\n");
2927 return i;
2928 }
2929
2930 i += scnprintf(buf, sz, "link data: %x", l->addr);
2931 i += scnprintf(buf + i, sz - i, " %x", l->state);
2932 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2933 i += scnprintf(buf + i, sz - i, " %u", l->session);
2934 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2935 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2936 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2937 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2938 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2939 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2940 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2941 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2942 i += scnprintf(buf + i, sz - i, " %u", 0);
2943 i += scnprintf(buf + i, sz - i, " %u", 0);
2944 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2945
2946 list = &l->transmq;
2947 len = skb_queue_len(list);
2948 hskb = skb_peek(list);
2949 tskb = skb_peek_tail(list);
2950 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2951 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2952 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2953
2954 list = &l->deferdq;
2955 len = skb_queue_len(list);
2956 hskb = skb_peek(list);
2957 tskb = skb_peek_tail(list);
2958 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2959 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2960 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2961
2962 list = &l->backlogq;
2963 len = skb_queue_len(list);
2964 hskb = skb_peek(list);
2965 tskb = skb_peek_tail(list);
2966 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2967 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2968 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2969
2970 list = l->inputq;
2971 len = skb_queue_len(list);
2972 hskb = skb_peek(list);
2973 tskb = skb_peek_tail(list);
2974 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2975 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2976 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2977
2978 if (dqueues & TIPC_DUMP_TRANSMQ) {
2979 i += scnprintf(buf + i, sz - i, "transmq: ");
2980 i += tipc_list_dump(&l->transmq, false, buf + i);
2981 }
2982 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2983 i += scnprintf(buf + i, sz - i,
2984 "backlogq: <%u %u %u %u %u>, ",
2985 l->backlog[TIPC_LOW_IMPORTANCE].len,
2986 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2987 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2988 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2989 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2990 i += tipc_list_dump(&l->backlogq, false, buf + i);
2991 }
2992 if (dqueues & TIPC_DUMP_DEFERDQ) {
2993 i += scnprintf(buf + i, sz - i, "deferdq: ");
2994 i += tipc_list_dump(&l->deferdq, false, buf + i);
2995 }
2996 if (dqueues & TIPC_DUMP_INPUTQ) {
2997 i += scnprintf(buf + i, sz - i, "inputq: ");
2998 i += tipc_list_dump(l->inputq, false, buf + i);
2999 }
3000 if (dqueues & TIPC_DUMP_WAKEUP) {
3001 i += scnprintf(buf + i, sz - i, "wakeup: ");
3002 i += tipc_list_dump(&l->wakeupq, false, buf + i);
3003 }
3004
3005 return i;
3006}
1/*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
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.
19 *
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.
23 *
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.
35 */
36
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
48#include <linux/pkt_sched.h>
49
50struct tipc_stats {
51 u32 sent_pkts;
52 u32 recv_pkts;
53 u32 sent_states;
54 u32 recv_states;
55 u32 sent_probes;
56 u32 recv_probes;
57 u32 sent_nacks;
58 u32 recv_nacks;
59 u32 sent_acks;
60 u32 sent_bundled;
61 u32 sent_bundles;
62 u32 recv_bundled;
63 u32 recv_bundles;
64 u32 retransmitted;
65 u32 sent_fragmented;
66 u32 sent_fragments;
67 u32 recv_fragmented;
68 u32 recv_fragments;
69 u32 link_congs; /* # port sends blocked by congestion */
70 u32 deferred_recv;
71 u32 duplicates;
72 u32 max_queue_sz; /* send queue size high water mark */
73 u32 accu_queue_sz; /* used for send queue size profiling */
74 u32 queue_sz_counts; /* used for send queue size profiling */
75 u32 msg_length_counts; /* used for message length profiling */
76 u32 msg_lengths_total; /* used for message length profiling */
77 u32 msg_length_profile[7]; /* used for msg. length profiling */
78};
79
80/**
81 * struct tipc_link - TIPC link data structure
82 * @addr: network address of link's peer node
83 * @name: link name character string
84 * @media_addr: media address to use when sending messages over link
85 * @timer: link timer
86 * @net: pointer to namespace struct
87 * @refcnt: reference counter for permanent references (owner node & timer)
88 * @peer_session: link session # being used by peer end of link
89 * @peer_bearer_id: bearer id used by link's peer endpoint
90 * @bearer_id: local bearer id used by link
91 * @tolerance: minimum link continuity loss needed to reset link [in ms]
92 * @abort_limit: # of unacknowledged continuity probes needed to reset link
93 * @state: current state of link FSM
94 * @peer_caps: bitmap describing capabilities of peer node
95 * @silent_intv_cnt: # of timer intervals without any reception from peer
96 * @proto_msg: template for control messages generated by link
97 * @pmsg: convenience pointer to "proto_msg" field
98 * @priority: current link priority
99 * @net_plane: current link network plane ('A' through 'H')
100 * @mon_state: cookie with information needed by link monitor
101 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
102 * @exp_msg_count: # of tunnelled messages expected during link changeover
103 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
104 * @mtu: current maximum packet size for this link
105 * @advertised_mtu: advertised own mtu when link is being established
106 * @transmitq: queue for sent, non-acked messages
107 * @backlogq: queue for messages waiting to be sent
108 * @snt_nxt: next sequence number to use for outbound messages
109 * @ackers: # of peers that needs to ack each packet before it can be released
110 * @acked: # last packet acked by a certain peer. Used for broadcast.
111 * @rcv_nxt: next sequence number to expect for inbound messages
112 * @deferred_queue: deferred queue saved OOS b'cast message received from node
113 * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
114 * @inputq: buffer queue for messages to be delivered upwards
115 * @namedq: buffer queue for name table messages to be delivered upwards
116 * @next_out: ptr to first unsent outbound message in queue
117 * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
118 * @long_msg_seq_no: next identifier to use for outbound fragmented messages
119 * @reasm_buf: head of partially reassembled inbound message fragments
120 * @bc_rcvr: marks that this is a broadcast receiver link
121 * @stats: collects statistics regarding link activity
122 */
123struct tipc_link {
124 u32 addr;
125 char name[TIPC_MAX_LINK_NAME];
126 struct net *net;
127
128 /* Management and link supervision data */
129 u16 peer_session;
130 u16 session;
131 u16 snd_nxt_state;
132 u16 rcv_nxt_state;
133 u32 peer_bearer_id;
134 u32 bearer_id;
135 u32 tolerance;
136 u32 abort_limit;
137 u32 state;
138 u16 peer_caps;
139 bool in_session;
140 bool active;
141 u32 silent_intv_cnt;
142 char if_name[TIPC_MAX_IF_NAME];
143 u32 priority;
144 char net_plane;
145 struct tipc_mon_state mon_state;
146 u16 rst_cnt;
147
148 /* Failover/synch */
149 u16 drop_point;
150 struct sk_buff *failover_reasm_skb;
151 struct sk_buff_head failover_deferdq;
152
153 /* Max packet negotiation */
154 u16 mtu;
155 u16 advertised_mtu;
156
157 /* Sending */
158 struct sk_buff_head transmq;
159 struct sk_buff_head backlogq;
160 struct {
161 u16 len;
162 u16 limit;
163 struct sk_buff *target_bskb;
164 } backlog[5];
165 u16 snd_nxt;
166 u16 window;
167
168 /* Reception */
169 u16 rcv_nxt;
170 u32 rcv_unacked;
171 struct sk_buff_head deferdq;
172 struct sk_buff_head *inputq;
173 struct sk_buff_head *namedq;
174
175 /* Congestion handling */
176 struct sk_buff_head wakeupq;
177
178 /* Fragmentation/reassembly */
179 struct sk_buff *reasm_buf;
180 struct sk_buff *reasm_tnlmsg;
181
182 /* Broadcast */
183 u16 ackers;
184 u16 acked;
185 struct tipc_link *bc_rcvlink;
186 struct tipc_link *bc_sndlink;
187 u8 nack_state;
188 bool bc_peer_is_up;
189
190 /* Statistics */
191 struct tipc_stats stats;
192};
193
194/*
195 * Error message prefixes
196 */
197static const char *link_co_err = "Link tunneling error, ";
198static const char *link_rst_msg = "Resetting link ";
199
200/* Send states for broadcast NACKs
201 */
202enum {
203 BC_NACK_SND_CONDITIONAL,
204 BC_NACK_SND_UNCONDITIONAL,
205 BC_NACK_SND_SUPPRESS,
206};
207
208#define TIPC_BC_RETR_LIM (jiffies + msecs_to_jiffies(10))
209#define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
210
211/*
212 * Interval between NACKs when packets arrive out of order
213 */
214#define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
215
216/* Link FSM states:
217 */
218enum {
219 LINK_ESTABLISHED = 0xe,
220 LINK_ESTABLISHING = 0xe << 4,
221 LINK_RESET = 0x1 << 8,
222 LINK_RESETTING = 0x2 << 12,
223 LINK_PEER_RESET = 0xd << 16,
224 LINK_FAILINGOVER = 0xf << 20,
225 LINK_SYNCHING = 0xc << 24
226};
227
228/* Link FSM state checking routines
229 */
230static int link_is_up(struct tipc_link *l)
231{
232 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
233}
234
235static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
236 struct sk_buff_head *xmitq);
237static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
238 bool probe_reply, u16 rcvgap,
239 int tolerance, int priority,
240 struct sk_buff_head *xmitq);
241static void link_print(struct tipc_link *l, const char *str);
242static int tipc_link_build_nack_msg(struct tipc_link *l,
243 struct sk_buff_head *xmitq);
244static void tipc_link_build_bc_init_msg(struct tipc_link *l,
245 struct sk_buff_head *xmitq);
246static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data);
248static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
249 struct tipc_gap_ack_blks *ga,
250 struct sk_buff_head *xmitq);
251
252/*
253 * Simple non-static link routines (i.e. referenced outside this file)
254 */
255bool tipc_link_is_up(struct tipc_link *l)
256{
257 return link_is_up(l);
258}
259
260bool tipc_link_peer_is_down(struct tipc_link *l)
261{
262 return l->state == LINK_PEER_RESET;
263}
264
265bool tipc_link_is_reset(struct tipc_link *l)
266{
267 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
268}
269
270bool tipc_link_is_establishing(struct tipc_link *l)
271{
272 return l->state == LINK_ESTABLISHING;
273}
274
275bool tipc_link_is_synching(struct tipc_link *l)
276{
277 return l->state == LINK_SYNCHING;
278}
279
280bool tipc_link_is_failingover(struct tipc_link *l)
281{
282 return l->state == LINK_FAILINGOVER;
283}
284
285bool tipc_link_is_blocked(struct tipc_link *l)
286{
287 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
288}
289
290static bool link_is_bc_sndlink(struct tipc_link *l)
291{
292 return !l->bc_sndlink;
293}
294
295static bool link_is_bc_rcvlink(struct tipc_link *l)
296{
297 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
298}
299
300void tipc_link_set_active(struct tipc_link *l, bool active)
301{
302 l->active = active;
303}
304
305u32 tipc_link_id(struct tipc_link *l)
306{
307 return l->peer_bearer_id << 16 | l->bearer_id;
308}
309
310int tipc_link_window(struct tipc_link *l)
311{
312 return l->window;
313}
314
315int tipc_link_prio(struct tipc_link *l)
316{
317 return l->priority;
318}
319
320unsigned long tipc_link_tolerance(struct tipc_link *l)
321{
322 return l->tolerance;
323}
324
325struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
326{
327 return l->inputq;
328}
329
330char tipc_link_plane(struct tipc_link *l)
331{
332 return l->net_plane;
333}
334
335void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
336{
337 l->peer_caps = capabilities;
338}
339
340void tipc_link_add_bc_peer(struct tipc_link *snd_l,
341 struct tipc_link *uc_l,
342 struct sk_buff_head *xmitq)
343{
344 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
345
346 snd_l->ackers++;
347 rcv_l->acked = snd_l->snd_nxt - 1;
348 snd_l->state = LINK_ESTABLISHED;
349 tipc_link_build_bc_init_msg(uc_l, xmitq);
350}
351
352void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
353 struct tipc_link *rcv_l,
354 struct sk_buff_head *xmitq)
355{
356 u16 ack = snd_l->snd_nxt - 1;
357
358 snd_l->ackers--;
359 rcv_l->bc_peer_is_up = true;
360 rcv_l->state = LINK_ESTABLISHED;
361 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
362 trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
363 tipc_link_reset(rcv_l);
364 rcv_l->state = LINK_RESET;
365 if (!snd_l->ackers) {
366 trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
367 tipc_link_reset(snd_l);
368 snd_l->state = LINK_RESET;
369 __skb_queue_purge(xmitq);
370 }
371}
372
373int tipc_link_bc_peers(struct tipc_link *l)
374{
375 return l->ackers;
376}
377
378static u16 link_bc_rcv_gap(struct tipc_link *l)
379{
380 struct sk_buff *skb = skb_peek(&l->deferdq);
381 u16 gap = 0;
382
383 if (more(l->snd_nxt, l->rcv_nxt))
384 gap = l->snd_nxt - l->rcv_nxt;
385 if (skb)
386 gap = buf_seqno(skb) - l->rcv_nxt;
387 return gap;
388}
389
390void tipc_link_set_mtu(struct tipc_link *l, int mtu)
391{
392 l->mtu = mtu;
393}
394
395int tipc_link_mtu(struct tipc_link *l)
396{
397 return l->mtu;
398}
399
400u16 tipc_link_rcv_nxt(struct tipc_link *l)
401{
402 return l->rcv_nxt;
403}
404
405u16 tipc_link_acked(struct tipc_link *l)
406{
407 return l->acked;
408}
409
410char *tipc_link_name(struct tipc_link *l)
411{
412 return l->name;
413}
414
415u32 tipc_link_state(struct tipc_link *l)
416{
417 return l->state;
418}
419
420/**
421 * tipc_link_create - create a new link
422 * @n: pointer to associated node
423 * @if_name: associated interface name
424 * @bearer_id: id (index) of associated bearer
425 * @tolerance: link tolerance to be used by link
426 * @net_plane: network plane (A,B,c..) this link belongs to
427 * @mtu: mtu to be advertised by link
428 * @priority: priority to be used by link
429 * @window: send window to be used by link
430 * @session: session to be used by link
431 * @ownnode: identity of own node
432 * @peer: node id of peer node
433 * @peer_caps: bitmap describing peer node capabilities
434 * @bc_sndlink: the namespace global link used for broadcast sending
435 * @bc_rcvlink: the peer specific link used for broadcast reception
436 * @inputq: queue to put messages ready for delivery
437 * @namedq: queue to put binding table update messages ready for delivery
438 * @link: return value, pointer to put the created link
439 *
440 * Returns true if link was created, otherwise false
441 */
442bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
443 int tolerance, char net_plane, u32 mtu, int priority,
444 int window, u32 session, u32 self,
445 u32 peer, u8 *peer_id, u16 peer_caps,
446 struct tipc_link *bc_sndlink,
447 struct tipc_link *bc_rcvlink,
448 struct sk_buff_head *inputq,
449 struct sk_buff_head *namedq,
450 struct tipc_link **link)
451{
452 char peer_str[NODE_ID_STR_LEN] = {0,};
453 char self_str[NODE_ID_STR_LEN] = {0,};
454 struct tipc_link *l;
455
456 l = kzalloc(sizeof(*l), GFP_ATOMIC);
457 if (!l)
458 return false;
459 *link = l;
460 l->session = session;
461
462 /* Set link name for unicast links only */
463 if (peer_id) {
464 tipc_nodeid2string(self_str, tipc_own_id(net));
465 if (strlen(self_str) > 16)
466 sprintf(self_str, "%x", self);
467 tipc_nodeid2string(peer_str, peer_id);
468 if (strlen(peer_str) > 16)
469 sprintf(peer_str, "%x", peer);
470 }
471 /* Peer i/f name will be completed by reset/activate message */
472 snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
473 self_str, if_name, peer_str);
474
475 strcpy(l->if_name, if_name);
476 l->addr = peer;
477 l->peer_caps = peer_caps;
478 l->net = net;
479 l->in_session = false;
480 l->bearer_id = bearer_id;
481 l->tolerance = tolerance;
482 if (bc_rcvlink)
483 bc_rcvlink->tolerance = tolerance;
484 l->net_plane = net_plane;
485 l->advertised_mtu = mtu;
486 l->mtu = mtu;
487 l->priority = priority;
488 tipc_link_set_queue_limits(l, window);
489 l->ackers = 1;
490 l->bc_sndlink = bc_sndlink;
491 l->bc_rcvlink = bc_rcvlink;
492 l->inputq = inputq;
493 l->namedq = namedq;
494 l->state = LINK_RESETTING;
495 __skb_queue_head_init(&l->transmq);
496 __skb_queue_head_init(&l->backlogq);
497 __skb_queue_head_init(&l->deferdq);
498 __skb_queue_head_init(&l->failover_deferdq);
499 skb_queue_head_init(&l->wakeupq);
500 skb_queue_head_init(l->inputq);
501 return true;
502}
503
504/**
505 * tipc_link_bc_create - create new link to be used for broadcast
506 * @n: pointer to associated node
507 * @mtu: mtu to be used initially if no peers
508 * @window: send window to be used
509 * @inputq: queue to put messages ready for delivery
510 * @namedq: queue to put binding table update messages ready for delivery
511 * @link: return value, pointer to put the created link
512 *
513 * Returns true if link was created, otherwise false
514 */
515bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
516 int mtu, int window, u16 peer_caps,
517 struct sk_buff_head *inputq,
518 struct sk_buff_head *namedq,
519 struct tipc_link *bc_sndlink,
520 struct tipc_link **link)
521{
522 struct tipc_link *l;
523
524 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
525 0, ownnode, peer, NULL, peer_caps, bc_sndlink,
526 NULL, inputq, namedq, link))
527 return false;
528
529 l = *link;
530 strcpy(l->name, tipc_bclink_name);
531 trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
532 tipc_link_reset(l);
533 l->state = LINK_RESET;
534 l->ackers = 0;
535 l->bc_rcvlink = l;
536
537 /* Broadcast send link is always up */
538 if (link_is_bc_sndlink(l))
539 l->state = LINK_ESTABLISHED;
540
541 /* Disable replicast if even a single peer doesn't support it */
542 if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
543 tipc_bcast_disable_rcast(net);
544
545 return true;
546}
547
548/**
549 * tipc_link_fsm_evt - link finite state machine
550 * @l: pointer to link
551 * @evt: state machine event to be processed
552 */
553int tipc_link_fsm_evt(struct tipc_link *l, int evt)
554{
555 int rc = 0;
556 int old_state = l->state;
557
558 switch (l->state) {
559 case LINK_RESETTING:
560 switch (evt) {
561 case LINK_PEER_RESET_EVT:
562 l->state = LINK_PEER_RESET;
563 break;
564 case LINK_RESET_EVT:
565 l->state = LINK_RESET;
566 break;
567 case LINK_FAILURE_EVT:
568 case LINK_FAILOVER_BEGIN_EVT:
569 case LINK_ESTABLISH_EVT:
570 case LINK_FAILOVER_END_EVT:
571 case LINK_SYNCH_BEGIN_EVT:
572 case LINK_SYNCH_END_EVT:
573 default:
574 goto illegal_evt;
575 }
576 break;
577 case LINK_RESET:
578 switch (evt) {
579 case LINK_PEER_RESET_EVT:
580 l->state = LINK_ESTABLISHING;
581 break;
582 case LINK_FAILOVER_BEGIN_EVT:
583 l->state = LINK_FAILINGOVER;
584 case LINK_FAILURE_EVT:
585 case LINK_RESET_EVT:
586 case LINK_ESTABLISH_EVT:
587 case LINK_FAILOVER_END_EVT:
588 break;
589 case LINK_SYNCH_BEGIN_EVT:
590 case LINK_SYNCH_END_EVT:
591 default:
592 goto illegal_evt;
593 }
594 break;
595 case LINK_PEER_RESET:
596 switch (evt) {
597 case LINK_RESET_EVT:
598 l->state = LINK_ESTABLISHING;
599 break;
600 case LINK_PEER_RESET_EVT:
601 case LINK_ESTABLISH_EVT:
602 case LINK_FAILURE_EVT:
603 break;
604 case LINK_SYNCH_BEGIN_EVT:
605 case LINK_SYNCH_END_EVT:
606 case LINK_FAILOVER_BEGIN_EVT:
607 case LINK_FAILOVER_END_EVT:
608 default:
609 goto illegal_evt;
610 }
611 break;
612 case LINK_FAILINGOVER:
613 switch (evt) {
614 case LINK_FAILOVER_END_EVT:
615 l->state = LINK_RESET;
616 break;
617 case LINK_PEER_RESET_EVT:
618 case LINK_RESET_EVT:
619 case LINK_ESTABLISH_EVT:
620 case LINK_FAILURE_EVT:
621 break;
622 case LINK_FAILOVER_BEGIN_EVT:
623 case LINK_SYNCH_BEGIN_EVT:
624 case LINK_SYNCH_END_EVT:
625 default:
626 goto illegal_evt;
627 }
628 break;
629 case LINK_ESTABLISHING:
630 switch (evt) {
631 case LINK_ESTABLISH_EVT:
632 l->state = LINK_ESTABLISHED;
633 break;
634 case LINK_FAILOVER_BEGIN_EVT:
635 l->state = LINK_FAILINGOVER;
636 break;
637 case LINK_RESET_EVT:
638 l->state = LINK_RESET;
639 break;
640 case LINK_FAILURE_EVT:
641 case LINK_PEER_RESET_EVT:
642 case LINK_SYNCH_BEGIN_EVT:
643 case LINK_FAILOVER_END_EVT:
644 break;
645 case LINK_SYNCH_END_EVT:
646 default:
647 goto illegal_evt;
648 }
649 break;
650 case LINK_ESTABLISHED:
651 switch (evt) {
652 case LINK_PEER_RESET_EVT:
653 l->state = LINK_PEER_RESET;
654 rc |= TIPC_LINK_DOWN_EVT;
655 break;
656 case LINK_FAILURE_EVT:
657 l->state = LINK_RESETTING;
658 rc |= TIPC_LINK_DOWN_EVT;
659 break;
660 case LINK_RESET_EVT:
661 l->state = LINK_RESET;
662 break;
663 case LINK_ESTABLISH_EVT:
664 case LINK_SYNCH_END_EVT:
665 break;
666 case LINK_SYNCH_BEGIN_EVT:
667 l->state = LINK_SYNCHING;
668 break;
669 case LINK_FAILOVER_BEGIN_EVT:
670 case LINK_FAILOVER_END_EVT:
671 default:
672 goto illegal_evt;
673 }
674 break;
675 case LINK_SYNCHING:
676 switch (evt) {
677 case LINK_PEER_RESET_EVT:
678 l->state = LINK_PEER_RESET;
679 rc |= TIPC_LINK_DOWN_EVT;
680 break;
681 case LINK_FAILURE_EVT:
682 l->state = LINK_RESETTING;
683 rc |= TIPC_LINK_DOWN_EVT;
684 break;
685 case LINK_RESET_EVT:
686 l->state = LINK_RESET;
687 break;
688 case LINK_ESTABLISH_EVT:
689 case LINK_SYNCH_BEGIN_EVT:
690 break;
691 case LINK_SYNCH_END_EVT:
692 l->state = LINK_ESTABLISHED;
693 break;
694 case LINK_FAILOVER_BEGIN_EVT:
695 case LINK_FAILOVER_END_EVT:
696 default:
697 goto illegal_evt;
698 }
699 break;
700 default:
701 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
702 }
703 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
704 return rc;
705illegal_evt:
706 pr_err("Illegal FSM event %x in state %x on link %s\n",
707 evt, l->state, l->name);
708 trace_tipc_link_fsm(l->name, old_state, l->state, evt);
709 return rc;
710}
711
712/* link_profile_stats - update statistical profiling of traffic
713 */
714static void link_profile_stats(struct tipc_link *l)
715{
716 struct sk_buff *skb;
717 struct tipc_msg *msg;
718 int length;
719
720 /* Update counters used in statistical profiling of send traffic */
721 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
722 l->stats.queue_sz_counts++;
723
724 skb = skb_peek(&l->transmq);
725 if (!skb)
726 return;
727 msg = buf_msg(skb);
728 length = msg_size(msg);
729
730 if (msg_user(msg) == MSG_FRAGMENTER) {
731 if (msg_type(msg) != FIRST_FRAGMENT)
732 return;
733 length = msg_size(msg_inner_hdr(msg));
734 }
735 l->stats.msg_lengths_total += length;
736 l->stats.msg_length_counts++;
737 if (length <= 64)
738 l->stats.msg_length_profile[0]++;
739 else if (length <= 256)
740 l->stats.msg_length_profile[1]++;
741 else if (length <= 1024)
742 l->stats.msg_length_profile[2]++;
743 else if (length <= 4096)
744 l->stats.msg_length_profile[3]++;
745 else if (length <= 16384)
746 l->stats.msg_length_profile[4]++;
747 else if (length <= 32768)
748 l->stats.msg_length_profile[5]++;
749 else
750 l->stats.msg_length_profile[6]++;
751}
752
753/**
754 * tipc_link_too_silent - check if link is "too silent"
755 * @l: tipc link to be checked
756 *
757 * Returns true if the link 'silent_intv_cnt' is about to reach the
758 * 'abort_limit' value, otherwise false
759 */
760bool tipc_link_too_silent(struct tipc_link *l)
761{
762 return (l->silent_intv_cnt + 2 > l->abort_limit);
763}
764
765/* tipc_link_timeout - perform periodic task as instructed from node timeout
766 */
767int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
768{
769 int mtyp = 0;
770 int rc = 0;
771 bool state = false;
772 bool probe = false;
773 bool setup = false;
774 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
775 u16 bc_acked = l->bc_rcvlink->acked;
776 struct tipc_mon_state *mstate = &l->mon_state;
777
778 trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
779 trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
780 switch (l->state) {
781 case LINK_ESTABLISHED:
782 case LINK_SYNCHING:
783 mtyp = STATE_MSG;
784 link_profile_stats(l);
785 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
786 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
787 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
788 state = bc_acked != bc_snt;
789 state |= l->bc_rcvlink->rcv_unacked;
790 state |= l->rcv_unacked;
791 state |= !skb_queue_empty(&l->transmq);
792 state |= !skb_queue_empty(&l->deferdq);
793 probe = mstate->probing;
794 probe |= l->silent_intv_cnt;
795 if (probe || mstate->monitoring)
796 l->silent_intv_cnt++;
797 break;
798 case LINK_RESET:
799 setup = l->rst_cnt++ <= 4;
800 setup |= !(l->rst_cnt % 16);
801 mtyp = RESET_MSG;
802 break;
803 case LINK_ESTABLISHING:
804 setup = true;
805 mtyp = ACTIVATE_MSG;
806 break;
807 case LINK_PEER_RESET:
808 case LINK_RESETTING:
809 case LINK_FAILINGOVER:
810 break;
811 default:
812 break;
813 }
814
815 if (state || probe || setup)
816 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
817
818 return rc;
819}
820
821/**
822 * link_schedule_user - schedule a message sender for wakeup after congestion
823 * @l: congested link
824 * @hdr: header of message that is being sent
825 * Create pseudo msg to send back to user when congestion abates
826 */
827static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
828{
829 u32 dnode = tipc_own_addr(l->net);
830 u32 dport = msg_origport(hdr);
831 struct sk_buff *skb;
832
833 /* Create and schedule wakeup pseudo message */
834 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
835 dnode, l->addr, dport, 0, 0);
836 if (!skb)
837 return -ENOBUFS;
838 msg_set_dest_droppable(buf_msg(skb), true);
839 TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
840 skb_queue_tail(&l->wakeupq, skb);
841 l->stats.link_congs++;
842 trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
843 return -ELINKCONG;
844}
845
846/**
847 * link_prepare_wakeup - prepare users for wakeup after congestion
848 * @l: congested link
849 * Wake up a number of waiting users, as permitted by available space
850 * in the send queue
851 */
852static void link_prepare_wakeup(struct tipc_link *l)
853{
854 struct sk_buff_head *wakeupq = &l->wakeupq;
855 struct sk_buff_head *inputq = l->inputq;
856 struct sk_buff *skb, *tmp;
857 struct sk_buff_head tmpq;
858 int avail[5] = {0,};
859 int imp = 0;
860
861 __skb_queue_head_init(&tmpq);
862
863 for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
864 avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
865
866 skb_queue_walk_safe(wakeupq, skb, tmp) {
867 imp = TIPC_SKB_CB(skb)->chain_imp;
868 if (avail[imp] <= 0)
869 continue;
870 avail[imp]--;
871 __skb_unlink(skb, wakeupq);
872 __skb_queue_tail(&tmpq, skb);
873 }
874
875 spin_lock_bh(&inputq->lock);
876 skb_queue_splice_tail(&tmpq, inputq);
877 spin_unlock_bh(&inputq->lock);
878
879}
880
881void tipc_link_reset(struct tipc_link *l)
882{
883 struct sk_buff_head list;
884 u32 imp;
885
886 __skb_queue_head_init(&list);
887
888 l->in_session = false;
889 /* Force re-synch of peer session number before establishing */
890 l->peer_session--;
891 l->session++;
892 l->mtu = l->advertised_mtu;
893
894 spin_lock_bh(&l->wakeupq.lock);
895 skb_queue_splice_init(&l->wakeupq, &list);
896 spin_unlock_bh(&l->wakeupq.lock);
897
898 spin_lock_bh(&l->inputq->lock);
899 skb_queue_splice_init(&list, l->inputq);
900 spin_unlock_bh(&l->inputq->lock);
901
902 __skb_queue_purge(&l->transmq);
903 __skb_queue_purge(&l->deferdq);
904 __skb_queue_purge(&l->backlogq);
905 __skb_queue_purge(&l->failover_deferdq);
906 for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
907 l->backlog[imp].len = 0;
908 l->backlog[imp].target_bskb = NULL;
909 }
910 kfree_skb(l->reasm_buf);
911 kfree_skb(l->reasm_tnlmsg);
912 kfree_skb(l->failover_reasm_skb);
913 l->reasm_buf = NULL;
914 l->reasm_tnlmsg = NULL;
915 l->failover_reasm_skb = NULL;
916 l->rcv_unacked = 0;
917 l->snd_nxt = 1;
918 l->rcv_nxt = 1;
919 l->snd_nxt_state = 1;
920 l->rcv_nxt_state = 1;
921 l->acked = 0;
922 l->silent_intv_cnt = 0;
923 l->rst_cnt = 0;
924 l->bc_peer_is_up = false;
925 memset(&l->mon_state, 0, sizeof(l->mon_state));
926 tipc_link_reset_stats(l);
927}
928
929/**
930 * tipc_link_xmit(): enqueue buffer list according to queue situation
931 * @link: link to use
932 * @list: chain of buffers containing message
933 * @xmitq: returned list of packets to be sent by caller
934 *
935 * Consumes the buffer chain.
936 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
937 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
938 */
939int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
940 struct sk_buff_head *xmitq)
941{
942 struct tipc_msg *hdr = buf_msg(skb_peek(list));
943 unsigned int maxwin = l->window;
944 int imp = msg_importance(hdr);
945 unsigned int mtu = l->mtu;
946 u16 ack = l->rcv_nxt - 1;
947 u16 seqno = l->snd_nxt;
948 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
949 struct sk_buff_head *transmq = &l->transmq;
950 struct sk_buff_head *backlogq = &l->backlogq;
951 struct sk_buff *skb, *_skb, **tskb;
952 int pkt_cnt = skb_queue_len(list);
953 int rc = 0;
954
955 if (unlikely(msg_size(hdr) > mtu)) {
956 pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
957 skb_queue_len(list), msg_user(hdr),
958 msg_type(hdr), msg_size(hdr), mtu);
959 __skb_queue_purge(list);
960 return -EMSGSIZE;
961 }
962
963 /* Allow oversubscription of one data msg per source at congestion */
964 if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
965 if (imp == TIPC_SYSTEM_IMPORTANCE) {
966 pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
967 return -ENOBUFS;
968 }
969 rc = link_schedule_user(l, hdr);
970 }
971
972 if (pkt_cnt > 1) {
973 l->stats.sent_fragmented++;
974 l->stats.sent_fragments += pkt_cnt;
975 }
976
977 /* Prepare each packet for sending, and add to relevant queue: */
978 while (skb_queue_len(list)) {
979 skb = skb_peek(list);
980 hdr = buf_msg(skb);
981 msg_set_seqno(hdr, seqno);
982 msg_set_ack(hdr, ack);
983 msg_set_bcast_ack(hdr, bc_ack);
984
985 if (likely(skb_queue_len(transmq) < maxwin)) {
986 _skb = skb_clone(skb, GFP_ATOMIC);
987 if (!_skb) {
988 __skb_queue_purge(list);
989 return -ENOBUFS;
990 }
991 __skb_dequeue(list);
992 __skb_queue_tail(transmq, skb);
993 /* next retransmit attempt */
994 if (link_is_bc_sndlink(l))
995 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
996 __skb_queue_tail(xmitq, _skb);
997 TIPC_SKB_CB(skb)->ackers = l->ackers;
998 l->rcv_unacked = 0;
999 l->stats.sent_pkts++;
1000 seqno++;
1001 continue;
1002 }
1003 tskb = &l->backlog[imp].target_bskb;
1004 if (tipc_msg_bundle(*tskb, hdr, mtu)) {
1005 kfree_skb(__skb_dequeue(list));
1006 l->stats.sent_bundled++;
1007 continue;
1008 }
1009 if (tipc_msg_make_bundle(tskb, hdr, mtu, l->addr)) {
1010 kfree_skb(__skb_dequeue(list));
1011 __skb_queue_tail(backlogq, *tskb);
1012 l->backlog[imp].len++;
1013 l->stats.sent_bundled++;
1014 l->stats.sent_bundles++;
1015 continue;
1016 }
1017 l->backlog[imp].target_bskb = NULL;
1018 l->backlog[imp].len += skb_queue_len(list);
1019 skb_queue_splice_tail_init(list, backlogq);
1020 }
1021 l->snd_nxt = seqno;
1022 return rc;
1023}
1024
1025static void tipc_link_advance_backlog(struct tipc_link *l,
1026 struct sk_buff_head *xmitq)
1027{
1028 struct sk_buff *skb, *_skb;
1029 struct tipc_msg *hdr;
1030 u16 seqno = l->snd_nxt;
1031 u16 ack = l->rcv_nxt - 1;
1032 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1033 u32 imp;
1034
1035 while (skb_queue_len(&l->transmq) < l->window) {
1036 skb = skb_peek(&l->backlogq);
1037 if (!skb)
1038 break;
1039 _skb = skb_clone(skb, GFP_ATOMIC);
1040 if (!_skb)
1041 break;
1042 __skb_dequeue(&l->backlogq);
1043 hdr = buf_msg(skb);
1044 imp = msg_importance(hdr);
1045 l->backlog[imp].len--;
1046 if (unlikely(skb == l->backlog[imp].target_bskb))
1047 l->backlog[imp].target_bskb = NULL;
1048 __skb_queue_tail(&l->transmq, skb);
1049 /* next retransmit attempt */
1050 if (link_is_bc_sndlink(l))
1051 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1052
1053 __skb_queue_tail(xmitq, _skb);
1054 TIPC_SKB_CB(skb)->ackers = l->ackers;
1055 msg_set_seqno(hdr, seqno);
1056 msg_set_ack(hdr, ack);
1057 msg_set_bcast_ack(hdr, bc_ack);
1058 l->rcv_unacked = 0;
1059 l->stats.sent_pkts++;
1060 seqno++;
1061 }
1062 l->snd_nxt = seqno;
1063}
1064
1065/**
1066 * link_retransmit_failure() - Detect repeated retransmit failures
1067 * @l: tipc link sender
1068 * @r: tipc link receiver (= l in case of unicast)
1069 * @rc: returned code
1070 *
1071 * Return: true if the repeated retransmit failures happens, otherwise
1072 * false
1073 */
1074static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1075 int *rc)
1076{
1077 struct sk_buff *skb = skb_peek(&l->transmq);
1078 struct tipc_msg *hdr;
1079
1080 if (!skb)
1081 return false;
1082
1083 if (!TIPC_SKB_CB(skb)->retr_cnt)
1084 return false;
1085
1086 if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1087 msecs_to_jiffies(r->tolerance)))
1088 return false;
1089
1090 hdr = buf_msg(skb);
1091 if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1092 return false;
1093
1094 pr_warn("Retransmission failure on link <%s>\n", l->name);
1095 link_print(l, "State of link ");
1096 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1097 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1098 pr_info("sqno %u, prev: %x, dest: %x\n",
1099 msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1100 pr_info("retr_stamp %d, retr_cnt %d\n",
1101 jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1102 TIPC_SKB_CB(skb)->retr_cnt);
1103
1104 trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1105 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1106 trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1107
1108 if (link_is_bc_sndlink(l)) {
1109 r->state = LINK_RESET;
1110 *rc = TIPC_LINK_DOWN_EVT;
1111 } else {
1112 *rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1113 }
1114
1115 return true;
1116}
1117
1118/* tipc_link_bc_retrans() - retransmit zero or more packets
1119 * @l: the link to transmit on
1120 * @r: the receiving link ordering the retransmit. Same as l if unicast
1121 * @from: retransmit from (inclusive) this sequence number
1122 * @to: retransmit to (inclusive) this sequence number
1123 * xmitq: queue for accumulating the retransmitted packets
1124 */
1125static int tipc_link_bc_retrans(struct tipc_link *l, struct tipc_link *r,
1126 u16 from, u16 to, struct sk_buff_head *xmitq)
1127{
1128 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
1129 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1130 u16 ack = l->rcv_nxt - 1;
1131 struct tipc_msg *hdr;
1132 int rc = 0;
1133
1134 if (!skb)
1135 return 0;
1136 if (less(to, from))
1137 return 0;
1138
1139 trace_tipc_link_retrans(r, from, to, &l->transmq);
1140
1141 if (link_retransmit_failure(l, r, &rc))
1142 return rc;
1143
1144 skb_queue_walk(&l->transmq, skb) {
1145 hdr = buf_msg(skb);
1146 if (less(msg_seqno(hdr), from))
1147 continue;
1148 if (more(msg_seqno(hdr), to))
1149 break;
1150
1151 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1152 continue;
1153 TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
1154 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE, GFP_ATOMIC);
1155 if (!_skb)
1156 return 0;
1157 hdr = buf_msg(_skb);
1158 msg_set_ack(hdr, ack);
1159 msg_set_bcast_ack(hdr, bc_ack);
1160 _skb->priority = TC_PRIO_CONTROL;
1161 __skb_queue_tail(xmitq, _skb);
1162 l->stats.retransmitted++;
1163
1164 /* Increase actual retrans counter & mark first time */
1165 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1166 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1167 }
1168 return 0;
1169}
1170
1171/* tipc_data_input - deliver data and name distr msgs to upper layer
1172 *
1173 * Consumes buffer if message is of right type
1174 * Node lock must be held
1175 */
1176static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1177 struct sk_buff_head *inputq)
1178{
1179 struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1180 struct tipc_msg *hdr = buf_msg(skb);
1181
1182 switch (msg_user(hdr)) {
1183 case TIPC_LOW_IMPORTANCE:
1184 case TIPC_MEDIUM_IMPORTANCE:
1185 case TIPC_HIGH_IMPORTANCE:
1186 case TIPC_CRITICAL_IMPORTANCE:
1187 if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1188 skb_queue_tail(mc_inputq, skb);
1189 return true;
1190 }
1191 /* fall through */
1192 case CONN_MANAGER:
1193 skb_queue_tail(inputq, skb);
1194 return true;
1195 case GROUP_PROTOCOL:
1196 skb_queue_tail(mc_inputq, skb);
1197 return true;
1198 case NAME_DISTRIBUTOR:
1199 l->bc_rcvlink->state = LINK_ESTABLISHED;
1200 skb_queue_tail(l->namedq, skb);
1201 return true;
1202 case MSG_BUNDLER:
1203 case TUNNEL_PROTOCOL:
1204 case MSG_FRAGMENTER:
1205 case BCAST_PROTOCOL:
1206 return false;
1207 default:
1208 pr_warn("Dropping received illegal msg type\n");
1209 kfree_skb(skb);
1210 return true;
1211 };
1212}
1213
1214/* tipc_link_input - process packet that has passed link protocol check
1215 *
1216 * Consumes buffer
1217 */
1218static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1219 struct sk_buff_head *inputq,
1220 struct sk_buff **reasm_skb)
1221{
1222 struct tipc_msg *hdr = buf_msg(skb);
1223 struct sk_buff *iskb;
1224 struct sk_buff_head tmpq;
1225 int usr = msg_user(hdr);
1226 int pos = 0;
1227
1228 if (usr == MSG_BUNDLER) {
1229 skb_queue_head_init(&tmpq);
1230 l->stats.recv_bundles++;
1231 l->stats.recv_bundled += msg_msgcnt(hdr);
1232 while (tipc_msg_extract(skb, &iskb, &pos))
1233 tipc_data_input(l, iskb, &tmpq);
1234 tipc_skb_queue_splice_tail(&tmpq, inputq);
1235 return 0;
1236 } else if (usr == MSG_FRAGMENTER) {
1237 l->stats.recv_fragments++;
1238 if (tipc_buf_append(reasm_skb, &skb)) {
1239 l->stats.recv_fragmented++;
1240 tipc_data_input(l, skb, inputq);
1241 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1242 pr_warn_ratelimited("Unable to build fragment list\n");
1243 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1244 }
1245 return 0;
1246 } else if (usr == BCAST_PROTOCOL) {
1247 tipc_bcast_lock(l->net);
1248 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1249 tipc_bcast_unlock(l->net);
1250 }
1251
1252 kfree_skb(skb);
1253 return 0;
1254}
1255
1256/* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1257 * inner message along with the ones in the old link's
1258 * deferdq
1259 * @l: tunnel link
1260 * @skb: TUNNEL_PROTOCOL message
1261 * @inputq: queue to put messages ready for delivery
1262 */
1263static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1264 struct sk_buff_head *inputq)
1265{
1266 struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1267 struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1268 struct sk_buff_head *fdefq = &l->failover_deferdq;
1269 struct tipc_msg *hdr = buf_msg(skb);
1270 struct sk_buff *iskb;
1271 int ipos = 0;
1272 int rc = 0;
1273 u16 seqno;
1274
1275 if (msg_type(hdr) == SYNCH_MSG) {
1276 kfree_skb(skb);
1277 return 0;
1278 }
1279
1280 /* Not a fragment? */
1281 if (likely(!msg_nof_fragms(hdr))) {
1282 if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1283 pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1284 skb_queue_len(fdefq));
1285 return 0;
1286 }
1287 kfree_skb(skb);
1288 } else {
1289 /* Set fragment type for buf_append */
1290 if (msg_fragm_no(hdr) == 1)
1291 msg_set_type(hdr, FIRST_FRAGMENT);
1292 else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1293 msg_set_type(hdr, FRAGMENT);
1294 else
1295 msg_set_type(hdr, LAST_FRAGMENT);
1296
1297 if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1298 /* Successful but non-complete reassembly? */
1299 if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1300 return 0;
1301 pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1302 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1303 }
1304 iskb = skb;
1305 }
1306
1307 do {
1308 seqno = buf_seqno(iskb);
1309 if (unlikely(less(seqno, l->drop_point))) {
1310 kfree_skb(iskb);
1311 continue;
1312 }
1313 if (unlikely(seqno != l->drop_point)) {
1314 __tipc_skb_queue_sorted(fdefq, seqno, iskb);
1315 continue;
1316 }
1317
1318 l->drop_point++;
1319 if (!tipc_data_input(l, iskb, inputq))
1320 rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1321 if (unlikely(rc))
1322 break;
1323 } while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1324
1325 return rc;
1326}
1327
1328static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1329{
1330 bool released = false;
1331 struct sk_buff *skb, *tmp;
1332
1333 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1334 if (more(buf_seqno(skb), acked))
1335 break;
1336 __skb_unlink(skb, &l->transmq);
1337 kfree_skb(skb);
1338 released = true;
1339 }
1340 return released;
1341}
1342
1343/* tipc_build_gap_ack_blks - build Gap ACK blocks
1344 * @l: tipc link that data have come with gaps in sequence if any
1345 * @data: data buffer to store the Gap ACK blocks after built
1346 *
1347 * returns the actual allocated memory size
1348 */
1349static u16 tipc_build_gap_ack_blks(struct tipc_link *l, void *data)
1350{
1351 struct sk_buff *skb = skb_peek(&l->deferdq);
1352 struct tipc_gap_ack_blks *ga = data;
1353 u16 len, expect, seqno = 0;
1354 u8 n = 0;
1355
1356 if (!skb)
1357 goto exit;
1358
1359 expect = buf_seqno(skb);
1360 skb_queue_walk(&l->deferdq, skb) {
1361 seqno = buf_seqno(skb);
1362 if (unlikely(more(seqno, expect))) {
1363 ga->gacks[n].ack = htons(expect - 1);
1364 ga->gacks[n].gap = htons(seqno - expect);
1365 if (++n >= MAX_GAP_ACK_BLKS) {
1366 pr_info_ratelimited("Too few Gap ACK blocks!\n");
1367 goto exit;
1368 }
1369 } else if (unlikely(less(seqno, expect))) {
1370 pr_warn("Unexpected skb in deferdq!\n");
1371 continue;
1372 }
1373 expect = seqno + 1;
1374 }
1375
1376 /* last block */
1377 ga->gacks[n].ack = htons(seqno);
1378 ga->gacks[n].gap = 0;
1379 n++;
1380
1381exit:
1382 len = tipc_gap_ack_blks_sz(n);
1383 ga->len = htons(len);
1384 ga->gack_cnt = n;
1385 return len;
1386}
1387
1388/* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1389 * acked packets, also doing retransmissions if
1390 * gaps found
1391 * @l: tipc link with transmq queue to be advanced
1392 * @acked: seqno of last packet acked by peer without any gaps before
1393 * @gap: # of gap packets
1394 * @ga: buffer pointer to Gap ACK blocks from peer
1395 * @xmitq: queue for accumulating the retransmitted packets if any
1396 *
1397 * In case of a repeated retransmit failures, the call will return shortly
1398 * with a returned code (e.g. TIPC_LINK_DOWN_EVT)
1399 */
1400static int tipc_link_advance_transmq(struct tipc_link *l, u16 acked, u16 gap,
1401 struct tipc_gap_ack_blks *ga,
1402 struct sk_buff_head *xmitq)
1403{
1404 struct sk_buff *skb, *_skb, *tmp;
1405 struct tipc_msg *hdr;
1406 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1407 u16 ack = l->rcv_nxt - 1;
1408 bool passed = false;
1409 u16 seqno, n = 0;
1410 int rc = 0;
1411
1412 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1413 seqno = buf_seqno(skb);
1414
1415next_gap_ack:
1416 if (less_eq(seqno, acked)) {
1417 /* release skb */
1418 __skb_unlink(skb, &l->transmq);
1419 kfree_skb(skb);
1420 } else if (less_eq(seqno, acked + gap)) {
1421 /* First, check if repeated retrans failures occurs? */
1422 if (!passed && link_retransmit_failure(l, l, &rc))
1423 return rc;
1424 passed = true;
1425
1426 /* retransmit skb if unrestricted*/
1427 if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1428 continue;
1429 TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
1430 _skb = __pskb_copy(skb, LL_MAX_HEADER + MIN_H_SIZE,
1431 GFP_ATOMIC);
1432 if (!_skb)
1433 continue;
1434 hdr = buf_msg(_skb);
1435 msg_set_ack(hdr, ack);
1436 msg_set_bcast_ack(hdr, bc_ack);
1437 _skb->priority = TC_PRIO_CONTROL;
1438 __skb_queue_tail(xmitq, _skb);
1439 l->stats.retransmitted++;
1440
1441 /* Increase actual retrans counter & mark first time */
1442 if (!TIPC_SKB_CB(skb)->retr_cnt++)
1443 TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1444 } else {
1445 /* retry with Gap ACK blocks if any */
1446 if (!ga || n >= ga->gack_cnt)
1447 break;
1448 acked = ntohs(ga->gacks[n].ack);
1449 gap = ntohs(ga->gacks[n].gap);
1450 n++;
1451 goto next_gap_ack;
1452 }
1453 }
1454
1455 return 0;
1456}
1457
1458/* tipc_link_build_state_msg: prepare link state message for transmission
1459 *
1460 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1461 * risk of ack storms towards the sender
1462 */
1463int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1464{
1465 if (!l)
1466 return 0;
1467
1468 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1469 if (link_is_bc_rcvlink(l)) {
1470 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1471 return 0;
1472 l->rcv_unacked = 0;
1473
1474 /* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1475 l->snd_nxt = l->rcv_nxt;
1476 return TIPC_LINK_SND_STATE;
1477 }
1478
1479 /* Unicast ACK */
1480 l->rcv_unacked = 0;
1481 l->stats.sent_acks++;
1482 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1483 return 0;
1484}
1485
1486/* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1487 */
1488void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1489{
1490 int mtyp = RESET_MSG;
1491 struct sk_buff *skb;
1492
1493 if (l->state == LINK_ESTABLISHING)
1494 mtyp = ACTIVATE_MSG;
1495
1496 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1497
1498 /* Inform peer that this endpoint is going down if applicable */
1499 skb = skb_peek_tail(xmitq);
1500 if (skb && (l->state == LINK_RESET))
1501 msg_set_peer_stopping(buf_msg(skb), 1);
1502}
1503
1504/* tipc_link_build_nack_msg: prepare link nack message for transmission
1505 * Note that sending of broadcast NACK is coordinated among nodes, to
1506 * reduce the risk of NACK storms towards the sender
1507 */
1508static int tipc_link_build_nack_msg(struct tipc_link *l,
1509 struct sk_buff_head *xmitq)
1510{
1511 u32 def_cnt = ++l->stats.deferred_recv;
1512 u32 defq_len = skb_queue_len(&l->deferdq);
1513 int match1, match2;
1514
1515 if (link_is_bc_rcvlink(l)) {
1516 match1 = def_cnt & 0xf;
1517 match2 = tipc_own_addr(l->net) & 0xf;
1518 if (match1 == match2)
1519 return TIPC_LINK_SND_STATE;
1520 return 0;
1521 }
1522
1523 if (defq_len >= 3 && !((defq_len - 3) % 16))
1524 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1525 return 0;
1526}
1527
1528/* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1529 * @l: the link that should handle the message
1530 * @skb: TIPC packet
1531 * @xmitq: queue to place packets to be sent after this call
1532 */
1533int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1534 struct sk_buff_head *xmitq)
1535{
1536 struct sk_buff_head *defq = &l->deferdq;
1537 struct tipc_msg *hdr = buf_msg(skb);
1538 u16 seqno, rcv_nxt, win_lim;
1539 int rc = 0;
1540
1541 /* Verify and update link state */
1542 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1543 return tipc_link_proto_rcv(l, skb, xmitq);
1544
1545 /* Don't send probe at next timeout expiration */
1546 l->silent_intv_cnt = 0;
1547
1548 do {
1549 hdr = buf_msg(skb);
1550 seqno = msg_seqno(hdr);
1551 rcv_nxt = l->rcv_nxt;
1552 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1553
1554 if (unlikely(!link_is_up(l))) {
1555 if (l->state == LINK_ESTABLISHING)
1556 rc = TIPC_LINK_UP_EVT;
1557 goto drop;
1558 }
1559
1560 /* Drop if outside receive window */
1561 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1562 l->stats.duplicates++;
1563 goto drop;
1564 }
1565
1566 /* Forward queues and wake up waiting users */
1567 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1568 tipc_link_advance_backlog(l, xmitq);
1569 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1570 link_prepare_wakeup(l);
1571 }
1572
1573 /* Defer delivery if sequence gap */
1574 if (unlikely(seqno != rcv_nxt)) {
1575 __tipc_skb_queue_sorted(defq, seqno, skb);
1576 rc |= tipc_link_build_nack_msg(l, xmitq);
1577 break;
1578 }
1579
1580 /* Deliver packet */
1581 l->rcv_nxt++;
1582 l->stats.recv_pkts++;
1583
1584 if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1585 rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1586 else if (!tipc_data_input(l, skb, l->inputq))
1587 rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1588 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1589 rc |= tipc_link_build_state_msg(l, xmitq);
1590 if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1591 break;
1592 } while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1593
1594 return rc;
1595drop:
1596 kfree_skb(skb);
1597 return rc;
1598}
1599
1600static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1601 bool probe_reply, u16 rcvgap,
1602 int tolerance, int priority,
1603 struct sk_buff_head *xmitq)
1604{
1605 struct tipc_link *bcl = l->bc_rcvlink;
1606 struct sk_buff *skb;
1607 struct tipc_msg *hdr;
1608 struct sk_buff_head *dfq = &l->deferdq;
1609 bool node_up = link_is_up(bcl);
1610 struct tipc_mon_state *mstate = &l->mon_state;
1611 int dlen = 0;
1612 void *data;
1613 u16 glen = 0;
1614
1615 /* Don't send protocol message during reset or link failover */
1616 if (tipc_link_is_blocked(l))
1617 return;
1618
1619 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1620 return;
1621
1622 if (!skb_queue_empty(dfq))
1623 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1624
1625 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1626 tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1627 l->addr, tipc_own_addr(l->net), 0, 0, 0);
1628 if (!skb)
1629 return;
1630
1631 hdr = buf_msg(skb);
1632 data = msg_data(hdr);
1633 msg_set_session(hdr, l->session);
1634 msg_set_bearer_id(hdr, l->bearer_id);
1635 msg_set_net_plane(hdr, l->net_plane);
1636 msg_set_next_sent(hdr, l->snd_nxt);
1637 msg_set_ack(hdr, l->rcv_nxt - 1);
1638 msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1639 msg_set_bc_ack_invalid(hdr, !node_up);
1640 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1641 msg_set_link_tolerance(hdr, tolerance);
1642 msg_set_linkprio(hdr, priority);
1643 msg_set_redundant_link(hdr, node_up);
1644 msg_set_seq_gap(hdr, 0);
1645 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1646
1647 if (mtyp == STATE_MSG) {
1648 if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1649 msg_set_seqno(hdr, l->snd_nxt_state++);
1650 msg_set_seq_gap(hdr, rcvgap);
1651 msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1652 msg_set_probe(hdr, probe);
1653 msg_set_is_keepalive(hdr, probe || probe_reply);
1654 if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1655 glen = tipc_build_gap_ack_blks(l, data);
1656 tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1657 msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1658 skb_trim(skb, INT_H_SIZE + glen + dlen);
1659 l->stats.sent_states++;
1660 l->rcv_unacked = 0;
1661 } else {
1662 /* RESET_MSG or ACTIVATE_MSG */
1663 if (mtyp == ACTIVATE_MSG) {
1664 msg_set_dest_session_valid(hdr, 1);
1665 msg_set_dest_session(hdr, l->peer_session);
1666 }
1667 msg_set_max_pkt(hdr, l->advertised_mtu);
1668 strcpy(data, l->if_name);
1669 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1670 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1671 }
1672 if (probe)
1673 l->stats.sent_probes++;
1674 if (rcvgap)
1675 l->stats.sent_nacks++;
1676 skb->priority = TC_PRIO_CONTROL;
1677 __skb_queue_tail(xmitq, skb);
1678 trace_tipc_proto_build(skb, false, l->name);
1679}
1680
1681void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1682 struct sk_buff_head *xmitq)
1683{
1684 u32 onode = tipc_own_addr(l->net);
1685 struct tipc_msg *hdr, *ihdr;
1686 struct sk_buff_head tnlq;
1687 struct sk_buff *skb;
1688 u32 dnode = l->addr;
1689
1690 __skb_queue_head_init(&tnlq);
1691 skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1692 INT_H_SIZE, BASIC_H_SIZE,
1693 dnode, onode, 0, 0, 0);
1694 if (!skb) {
1695 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1696 return;
1697 }
1698
1699 hdr = buf_msg(skb);
1700 msg_set_msgcnt(hdr, 1);
1701 msg_set_bearer_id(hdr, l->peer_bearer_id);
1702
1703 ihdr = (struct tipc_msg *)msg_data(hdr);
1704 tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1705 BASIC_H_SIZE, dnode);
1706 msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1707 __skb_queue_tail(&tnlq, skb);
1708 tipc_link_xmit(l, &tnlq, xmitq);
1709}
1710
1711/* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1712 * with contents of the link's transmit and backlog queues.
1713 */
1714void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1715 int mtyp, struct sk_buff_head *xmitq)
1716{
1717 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1718 struct sk_buff *skb, *tnlskb;
1719 struct tipc_msg *hdr, tnlhdr;
1720 struct sk_buff_head *queue = &l->transmq;
1721 struct sk_buff_head tmpxq, tnlq, frags;
1722 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1723 bool pktcnt_need_update = false;
1724 u16 syncpt;
1725 int rc;
1726
1727 if (!tnl)
1728 return;
1729
1730 __skb_queue_head_init(&tnlq);
1731 __skb_queue_head_init(&tmpxq);
1732 __skb_queue_head_init(&frags);
1733
1734 /* At least one packet required for safe algorithm => add dummy */
1735 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1736 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1737 0, 0, TIPC_ERR_NO_PORT);
1738 if (!skb) {
1739 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1740 return;
1741 }
1742 __skb_queue_tail(&tnlq, skb);
1743 tipc_link_xmit(l, &tnlq, &tmpxq);
1744 __skb_queue_purge(&tmpxq);
1745
1746 /* Link Synching:
1747 * From now on, send only one single ("dummy") SYNCH message
1748 * to peer. The SYNCH message does not contain any data, just
1749 * a header conveying the synch point to the peer.
1750 */
1751 if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1752 tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1753 INT_H_SIZE, 0, l->addr,
1754 tipc_own_addr(l->net),
1755 0, 0, 0);
1756 if (!tnlskb) {
1757 pr_warn("%sunable to create dummy SYNCH_MSG\n",
1758 link_co_err);
1759 return;
1760 }
1761
1762 hdr = buf_msg(tnlskb);
1763 syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1764 msg_set_syncpt(hdr, syncpt);
1765 msg_set_bearer_id(hdr, l->peer_bearer_id);
1766 __skb_queue_tail(&tnlq, tnlskb);
1767 tipc_link_xmit(tnl, &tnlq, xmitq);
1768 return;
1769 }
1770
1771 /* Initialize reusable tunnel packet header */
1772 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1773 mtyp, INT_H_SIZE, l->addr);
1774 if (mtyp == SYNCH_MSG)
1775 pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1776 else
1777 pktcnt = skb_queue_len(&l->transmq);
1778 pktcnt += skb_queue_len(&l->backlogq);
1779 msg_set_msgcnt(&tnlhdr, pktcnt);
1780 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1781tnl:
1782 /* Wrap each packet into a tunnel packet */
1783 skb_queue_walk(queue, skb) {
1784 hdr = buf_msg(skb);
1785 if (queue == &l->backlogq)
1786 msg_set_seqno(hdr, seqno++);
1787 pktlen = msg_size(hdr);
1788
1789 /* Tunnel link MTU is not large enough? This could be
1790 * due to:
1791 * 1) Link MTU has just changed or set differently;
1792 * 2) Or FAILOVER on the top of a SYNCH message
1793 *
1794 * The 2nd case should not happen if peer supports
1795 * TIPC_TUNNEL_ENHANCED
1796 */
1797 if (pktlen > tnl->mtu - INT_H_SIZE) {
1798 if (mtyp == FAILOVER_MSG &&
1799 (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1800 rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
1801 &frags);
1802 if (rc) {
1803 pr_warn("%sunable to frag msg: rc %d\n",
1804 link_co_err, rc);
1805 return;
1806 }
1807 pktcnt += skb_queue_len(&frags) - 1;
1808 pktcnt_need_update = true;
1809 skb_queue_splice_tail_init(&frags, &tnlq);
1810 continue;
1811 }
1812 /* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
1813 * => Just warn it and return!
1814 */
1815 pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
1816 link_co_err, msg_user(hdr),
1817 msg_type(hdr), msg_size(hdr));
1818 return;
1819 }
1820
1821 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1822 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
1823 if (!tnlskb) {
1824 pr_warn("%sunable to send packet\n", link_co_err);
1825 return;
1826 }
1827 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1828 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1829 __skb_queue_tail(&tnlq, tnlskb);
1830 }
1831 if (queue != &l->backlogq) {
1832 queue = &l->backlogq;
1833 goto tnl;
1834 }
1835
1836 if (pktcnt_need_update)
1837 skb_queue_walk(&tnlq, skb) {
1838 hdr = buf_msg(skb);
1839 msg_set_msgcnt(hdr, pktcnt);
1840 }
1841
1842 tipc_link_xmit(tnl, &tnlq, xmitq);
1843
1844 if (mtyp == FAILOVER_MSG) {
1845 tnl->drop_point = l->rcv_nxt;
1846 tnl->failover_reasm_skb = l->reasm_buf;
1847 l->reasm_buf = NULL;
1848
1849 /* Failover the link's deferdq */
1850 if (unlikely(!skb_queue_empty(fdefq))) {
1851 pr_warn("Link failover deferdq not empty: %d!\n",
1852 skb_queue_len(fdefq));
1853 __skb_queue_purge(fdefq);
1854 }
1855 skb_queue_splice_init(&l->deferdq, fdefq);
1856 }
1857}
1858
1859/**
1860 * tipc_link_failover_prepare() - prepare tnl for link failover
1861 *
1862 * This is a special version of the precursor - tipc_link_tnl_prepare(),
1863 * see the tipc_node_link_failover() for details
1864 *
1865 * @l: failover link
1866 * @tnl: tunnel link
1867 * @xmitq: queue for messages to be xmited
1868 */
1869void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
1870 struct sk_buff_head *xmitq)
1871{
1872 struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1873
1874 tipc_link_create_dummy_tnl_msg(tnl, xmitq);
1875
1876 /* This failover link enpoint was never established before,
1877 * so it has not received anything from peer.
1878 * Otherwise, it must be a normal failover situation or the
1879 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
1880 * would have to start over from scratch instead.
1881 */
1882 tnl->drop_point = 1;
1883 tnl->failover_reasm_skb = NULL;
1884
1885 /* Initiate the link's failover deferdq */
1886 if (unlikely(!skb_queue_empty(fdefq))) {
1887 pr_warn("Link failover deferdq not empty: %d!\n",
1888 skb_queue_len(fdefq));
1889 __skb_queue_purge(fdefq);
1890 }
1891}
1892
1893/* tipc_link_validate_msg(): validate message against current link state
1894 * Returns true if message should be accepted, otherwise false
1895 */
1896bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
1897{
1898 u16 curr_session = l->peer_session;
1899 u16 session = msg_session(hdr);
1900 int mtyp = msg_type(hdr);
1901
1902 if (msg_user(hdr) != LINK_PROTOCOL)
1903 return true;
1904
1905 switch (mtyp) {
1906 case RESET_MSG:
1907 if (!l->in_session)
1908 return true;
1909 /* Accept only RESET with new session number */
1910 return more(session, curr_session);
1911 case ACTIVATE_MSG:
1912 if (!l->in_session)
1913 return true;
1914 /* Accept only ACTIVATE with new or current session number */
1915 return !less(session, curr_session);
1916 case STATE_MSG:
1917 /* Accept only STATE with current session number */
1918 if (!l->in_session)
1919 return false;
1920 if (session != curr_session)
1921 return false;
1922 /* Extra sanity check */
1923 if (!link_is_up(l) && msg_ack(hdr))
1924 return false;
1925 if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
1926 return true;
1927 /* Accept only STATE with new sequence number */
1928 return !less(msg_seqno(hdr), l->rcv_nxt_state);
1929 default:
1930 return false;
1931 }
1932}
1933
1934/* tipc_link_proto_rcv(): receive link level protocol message :
1935 * Note that network plane id propagates through the network, and may
1936 * change at any time. The node with lowest numerical id determines
1937 * network plane
1938 */
1939static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1940 struct sk_buff_head *xmitq)
1941{
1942 struct tipc_msg *hdr = buf_msg(skb);
1943 struct tipc_gap_ack_blks *ga = NULL;
1944 u16 rcvgap = 0;
1945 u16 ack = msg_ack(hdr);
1946 u16 gap = msg_seq_gap(hdr);
1947 u16 peers_snd_nxt = msg_next_sent(hdr);
1948 u16 peers_tol = msg_link_tolerance(hdr);
1949 u16 peers_prio = msg_linkprio(hdr);
1950 u16 rcv_nxt = l->rcv_nxt;
1951 u16 dlen = msg_data_sz(hdr);
1952 int mtyp = msg_type(hdr);
1953 bool reply = msg_probe(hdr);
1954 u16 glen = 0;
1955 void *data;
1956 char *if_name;
1957 int rc = 0;
1958
1959 trace_tipc_proto_rcv(skb, false, l->name);
1960 if (tipc_link_is_blocked(l) || !xmitq)
1961 goto exit;
1962
1963 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1964 l->net_plane = msg_net_plane(hdr);
1965
1966 skb_linearize(skb);
1967 hdr = buf_msg(skb);
1968 data = msg_data(hdr);
1969
1970 if (!tipc_link_validate_msg(l, hdr)) {
1971 trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
1972 trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
1973 goto exit;
1974 }
1975
1976 switch (mtyp) {
1977 case RESET_MSG:
1978 case ACTIVATE_MSG:
1979 /* Complete own link name with peer's interface name */
1980 if_name = strrchr(l->name, ':') + 1;
1981 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1982 break;
1983 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1984 break;
1985 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1986
1987 /* Update own tolerance if peer indicates a non-zero value */
1988 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
1989 l->tolerance = peers_tol;
1990 l->bc_rcvlink->tolerance = peers_tol;
1991 }
1992 /* Update own priority if peer's priority is higher */
1993 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1994 l->priority = peers_prio;
1995
1996 /* If peer is going down we want full re-establish cycle */
1997 if (msg_peer_stopping(hdr)) {
1998 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1999 break;
2000 }
2001
2002 /* If this endpoint was re-created while peer was ESTABLISHING
2003 * it doesn't know current session number. Force re-synch.
2004 */
2005 if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2006 l->session != msg_dest_session(hdr)) {
2007 if (less(l->session, msg_dest_session(hdr)))
2008 l->session = msg_dest_session(hdr) + 1;
2009 break;
2010 }
2011
2012 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2013 if (mtyp == RESET_MSG || !link_is_up(l))
2014 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2015
2016 /* ACTIVATE_MSG takes up link if it was already locally reset */
2017 if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2018 rc = TIPC_LINK_UP_EVT;
2019
2020 l->peer_session = msg_session(hdr);
2021 l->in_session = true;
2022 l->peer_bearer_id = msg_bearer_id(hdr);
2023 if (l->mtu > msg_max_pkt(hdr))
2024 l->mtu = msg_max_pkt(hdr);
2025 break;
2026
2027 case STATE_MSG:
2028 l->rcv_nxt_state = msg_seqno(hdr) + 1;
2029
2030 /* Update own tolerance if peer indicates a non-zero value */
2031 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2032 l->tolerance = peers_tol;
2033 l->bc_rcvlink->tolerance = peers_tol;
2034 }
2035 /* Update own prio if peer indicates a different value */
2036 if ((peers_prio != l->priority) &&
2037 in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2038 l->priority = peers_prio;
2039 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2040 }
2041
2042 l->silent_intv_cnt = 0;
2043 l->stats.recv_states++;
2044 if (msg_probe(hdr))
2045 l->stats.recv_probes++;
2046
2047 if (!link_is_up(l)) {
2048 if (l->state == LINK_ESTABLISHING)
2049 rc = TIPC_LINK_UP_EVT;
2050 break;
2051 }
2052
2053 /* Receive Gap ACK blocks from peer if any */
2054 if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
2055 ga = (struct tipc_gap_ack_blks *)data;
2056 glen = ntohs(ga->len);
2057 /* sanity check: if failed, ignore Gap ACK blocks */
2058 if (glen != tipc_gap_ack_blks_sz(ga->gack_cnt))
2059 ga = NULL;
2060 }
2061
2062 tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2063 &l->mon_state, l->bearer_id);
2064
2065 /* Send NACK if peer has sent pkts we haven't received yet */
2066 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
2067 rcvgap = peers_snd_nxt - l->rcv_nxt;
2068 if (rcvgap || reply)
2069 tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2070 rcvgap, 0, 0, xmitq);
2071
2072 rc |= tipc_link_advance_transmq(l, ack, gap, ga, xmitq);
2073
2074 /* If NACK, retransmit will now start at right position */
2075 if (gap)
2076 l->stats.recv_nacks++;
2077
2078 tipc_link_advance_backlog(l, xmitq);
2079 if (unlikely(!skb_queue_empty(&l->wakeupq)))
2080 link_prepare_wakeup(l);
2081 }
2082exit:
2083 kfree_skb(skb);
2084 return rc;
2085}
2086
2087/* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2088 */
2089static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2090 u16 peers_snd_nxt,
2091 struct sk_buff_head *xmitq)
2092{
2093 struct sk_buff *skb;
2094 struct tipc_msg *hdr;
2095 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2096 u16 ack = l->rcv_nxt - 1;
2097 u16 gap_to = peers_snd_nxt - 1;
2098
2099 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2100 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2101 if (!skb)
2102 return false;
2103 hdr = buf_msg(skb);
2104 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2105 msg_set_bcast_ack(hdr, ack);
2106 msg_set_bcgap_after(hdr, ack);
2107 if (dfrd_skb)
2108 gap_to = buf_seqno(dfrd_skb) - 1;
2109 msg_set_bcgap_to(hdr, gap_to);
2110 msg_set_non_seq(hdr, bcast);
2111 __skb_queue_tail(xmitq, skb);
2112 return true;
2113}
2114
2115/* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2116 *
2117 * Give a newly added peer node the sequence number where it should
2118 * start receiving and acking broadcast packets.
2119 */
2120static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2121 struct sk_buff_head *xmitq)
2122{
2123 struct sk_buff_head list;
2124
2125 __skb_queue_head_init(&list);
2126 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2127 return;
2128 msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2129 tipc_link_xmit(l, &list, xmitq);
2130}
2131
2132/* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2133 */
2134void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2135{
2136 int mtyp = msg_type(hdr);
2137 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2138
2139 if (link_is_up(l))
2140 return;
2141
2142 if (msg_user(hdr) == BCAST_PROTOCOL) {
2143 l->rcv_nxt = peers_snd_nxt;
2144 l->state = LINK_ESTABLISHED;
2145 return;
2146 }
2147
2148 if (l->peer_caps & TIPC_BCAST_SYNCH)
2149 return;
2150
2151 if (msg_peer_node_is_up(hdr))
2152 return;
2153
2154 /* Compatibility: accept older, less safe initial synch data */
2155 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2156 l->rcv_nxt = peers_snd_nxt;
2157}
2158
2159/* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2160 */
2161int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2162 struct sk_buff_head *xmitq)
2163{
2164 struct tipc_link *snd_l = l->bc_sndlink;
2165 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2166 u16 from = msg_bcast_ack(hdr) + 1;
2167 u16 to = from + msg_bc_gap(hdr) - 1;
2168 int rc = 0;
2169
2170 if (!link_is_up(l))
2171 return rc;
2172
2173 if (!msg_peer_node_is_up(hdr))
2174 return rc;
2175
2176 /* Open when peer ackowledges our bcast init msg (pkt #1) */
2177 if (msg_ack(hdr))
2178 l->bc_peer_is_up = true;
2179
2180 if (!l->bc_peer_is_up)
2181 return rc;
2182
2183 l->stats.recv_nacks++;
2184
2185 /* Ignore if peers_snd_nxt goes beyond receive window */
2186 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2187 return rc;
2188
2189 rc = tipc_link_bc_retrans(snd_l, l, from, to, xmitq);
2190
2191 l->snd_nxt = peers_snd_nxt;
2192 if (link_bc_rcv_gap(l))
2193 rc |= TIPC_LINK_SND_STATE;
2194
2195 /* Return now if sender supports nack via STATE messages */
2196 if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2197 return rc;
2198
2199 /* Otherwise, be backwards compatible */
2200
2201 if (!more(peers_snd_nxt, l->rcv_nxt)) {
2202 l->nack_state = BC_NACK_SND_CONDITIONAL;
2203 return 0;
2204 }
2205
2206 /* Don't NACK if one was recently sent or peeked */
2207 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2208 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2209 return 0;
2210 }
2211
2212 /* Conditionally delay NACK sending until next synch rcv */
2213 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2214 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2215 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2216 return 0;
2217 }
2218
2219 /* Send NACK now but suppress next one */
2220 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2221 l->nack_state = BC_NACK_SND_SUPPRESS;
2222 return 0;
2223}
2224
2225void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
2226 struct sk_buff_head *xmitq)
2227{
2228 struct sk_buff *skb, *tmp;
2229 struct tipc_link *snd_l = l->bc_sndlink;
2230
2231 if (!link_is_up(l) || !l->bc_peer_is_up)
2232 return;
2233
2234 if (!more(acked, l->acked))
2235 return;
2236
2237 trace_tipc_link_bc_ack(l, l->acked, acked, &snd_l->transmq);
2238 /* Skip over packets peer has already acked */
2239 skb_queue_walk(&snd_l->transmq, skb) {
2240 if (more(buf_seqno(skb), l->acked))
2241 break;
2242 }
2243
2244 /* Update/release the packets peer is acking now */
2245 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
2246 if (more(buf_seqno(skb), acked))
2247 break;
2248 if (!--TIPC_SKB_CB(skb)->ackers) {
2249 __skb_unlink(skb, &snd_l->transmq);
2250 kfree_skb(skb);
2251 }
2252 }
2253 l->acked = acked;
2254 tipc_link_advance_backlog(snd_l, xmitq);
2255 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
2256 link_prepare_wakeup(snd_l);
2257}
2258
2259/* tipc_link_bc_nack_rcv(): receive broadcast nack message
2260 * This function is here for backwards compatibility, since
2261 * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2262 */
2263int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2264 struct sk_buff_head *xmitq)
2265{
2266 struct tipc_msg *hdr = buf_msg(skb);
2267 u32 dnode = msg_destnode(hdr);
2268 int mtyp = msg_type(hdr);
2269 u16 acked = msg_bcast_ack(hdr);
2270 u16 from = acked + 1;
2271 u16 to = msg_bcgap_to(hdr);
2272 u16 peers_snd_nxt = to + 1;
2273 int rc = 0;
2274
2275 kfree_skb(skb);
2276
2277 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2278 return 0;
2279
2280 if (mtyp != STATE_MSG)
2281 return 0;
2282
2283 if (dnode == tipc_own_addr(l->net)) {
2284 tipc_link_bc_ack_rcv(l, acked, xmitq);
2285 rc = tipc_link_bc_retrans(l->bc_sndlink, l, from, to, xmitq);
2286 l->stats.recv_nacks++;
2287 return rc;
2288 }
2289
2290 /* Msg for other node => suppress own NACK at next sync if applicable */
2291 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2292 l->nack_state = BC_NACK_SND_SUPPRESS;
2293
2294 return 0;
2295}
2296
2297void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
2298{
2299 int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2300
2301 l->window = win;
2302 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
2303 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
2304 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
2305 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
2306 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
2307}
2308
2309/**
2310 * link_reset_stats - reset link statistics
2311 * @l: pointer to link
2312 */
2313void tipc_link_reset_stats(struct tipc_link *l)
2314{
2315 memset(&l->stats, 0, sizeof(l->stats));
2316}
2317
2318static void link_print(struct tipc_link *l, const char *str)
2319{
2320 struct sk_buff *hskb = skb_peek(&l->transmq);
2321 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2322 u16 tail = l->snd_nxt - 1;
2323
2324 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2325 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2326 skb_queue_len(&l->transmq), head, tail,
2327 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2328}
2329
2330/* Parse and validate nested (link) properties valid for media, bearer and link
2331 */
2332int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2333{
2334 int err;
2335
2336 err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2337 tipc_nl_prop_policy, NULL);
2338 if (err)
2339 return err;
2340
2341 if (props[TIPC_NLA_PROP_PRIO]) {
2342 u32 prio;
2343
2344 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2345 if (prio > TIPC_MAX_LINK_PRI)
2346 return -EINVAL;
2347 }
2348
2349 if (props[TIPC_NLA_PROP_TOL]) {
2350 u32 tol;
2351
2352 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2353 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2354 return -EINVAL;
2355 }
2356
2357 if (props[TIPC_NLA_PROP_WIN]) {
2358 u32 win;
2359
2360 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2361 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
2362 return -EINVAL;
2363 }
2364
2365 return 0;
2366}
2367
2368static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2369{
2370 int i;
2371 struct nlattr *stats;
2372
2373 struct nla_map {
2374 u32 key;
2375 u32 val;
2376 };
2377
2378 struct nla_map map[] = {
2379 {TIPC_NLA_STATS_RX_INFO, 0},
2380 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2381 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2382 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2383 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2384 {TIPC_NLA_STATS_TX_INFO, 0},
2385 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2386 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2387 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2388 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2389 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2390 s->msg_length_counts : 1},
2391 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2392 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2393 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2394 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2395 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2396 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2397 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2398 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2399 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2400 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
2401 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2402 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2403 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2404 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
2405 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2406 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2407 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2408 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2409 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2410 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2411 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2412 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2413 (s->accu_queue_sz / s->queue_sz_counts) : 0}
2414 };
2415
2416 stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2417 if (!stats)
2418 return -EMSGSIZE;
2419
2420 for (i = 0; i < ARRAY_SIZE(map); i++)
2421 if (nla_put_u32(skb, map[i].key, map[i].val))
2422 goto msg_full;
2423
2424 nla_nest_end(skb, stats);
2425
2426 return 0;
2427msg_full:
2428 nla_nest_cancel(skb, stats);
2429
2430 return -EMSGSIZE;
2431}
2432
2433/* Caller should hold appropriate locks to protect the link */
2434int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2435 struct tipc_link *link, int nlflags)
2436{
2437 u32 self = tipc_own_addr(net);
2438 struct nlattr *attrs;
2439 struct nlattr *prop;
2440 void *hdr;
2441 int err;
2442
2443 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2444 nlflags, TIPC_NL_LINK_GET);
2445 if (!hdr)
2446 return -EMSGSIZE;
2447
2448 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2449 if (!attrs)
2450 goto msg_full;
2451
2452 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2453 goto attr_msg_full;
2454 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2455 goto attr_msg_full;
2456 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2457 goto attr_msg_full;
2458 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2459 goto attr_msg_full;
2460 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2461 goto attr_msg_full;
2462
2463 if (tipc_link_is_up(link))
2464 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2465 goto attr_msg_full;
2466 if (link->active)
2467 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2468 goto attr_msg_full;
2469
2470 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2471 if (!prop)
2472 goto attr_msg_full;
2473 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2474 goto prop_msg_full;
2475 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2476 goto prop_msg_full;
2477 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2478 link->window))
2479 goto prop_msg_full;
2480 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2481 goto prop_msg_full;
2482 nla_nest_end(msg->skb, prop);
2483
2484 err = __tipc_nl_add_stats(msg->skb, &link->stats);
2485 if (err)
2486 goto attr_msg_full;
2487
2488 nla_nest_end(msg->skb, attrs);
2489 genlmsg_end(msg->skb, hdr);
2490
2491 return 0;
2492
2493prop_msg_full:
2494 nla_nest_cancel(msg->skb, prop);
2495attr_msg_full:
2496 nla_nest_cancel(msg->skb, attrs);
2497msg_full:
2498 genlmsg_cancel(msg->skb, hdr);
2499
2500 return -EMSGSIZE;
2501}
2502
2503static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2504 struct tipc_stats *stats)
2505{
2506 int i;
2507 struct nlattr *nest;
2508
2509 struct nla_map {
2510 __u32 key;
2511 __u32 val;
2512 };
2513
2514 struct nla_map map[] = {
2515 {TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2516 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2517 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2518 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2519 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2520 {TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2521 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2522 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2523 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2524 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2525 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2526 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2527 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2528 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2529 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2530 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2531 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2532 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2533 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2534 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2535 };
2536
2537 nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2538 if (!nest)
2539 return -EMSGSIZE;
2540
2541 for (i = 0; i < ARRAY_SIZE(map); i++)
2542 if (nla_put_u32(skb, map[i].key, map[i].val))
2543 goto msg_full;
2544
2545 nla_nest_end(skb, nest);
2546
2547 return 0;
2548msg_full:
2549 nla_nest_cancel(skb, nest);
2550
2551 return -EMSGSIZE;
2552}
2553
2554int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
2555{
2556 int err;
2557 void *hdr;
2558 struct nlattr *attrs;
2559 struct nlattr *prop;
2560 struct tipc_net *tn = net_generic(net, tipc_net_id);
2561 u32 bc_mode = tipc_bcast_get_broadcast_mode(net);
2562 u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2563 struct tipc_link *bcl = tn->bcl;
2564
2565 if (!bcl)
2566 return 0;
2567
2568 tipc_bcast_lock(net);
2569
2570 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2571 NLM_F_MULTI, TIPC_NL_LINK_GET);
2572 if (!hdr) {
2573 tipc_bcast_unlock(net);
2574 return -EMSGSIZE;
2575 }
2576
2577 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2578 if (!attrs)
2579 goto msg_full;
2580
2581 /* The broadcast link is always up */
2582 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2583 goto attr_msg_full;
2584
2585 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2586 goto attr_msg_full;
2587 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2588 goto attr_msg_full;
2589 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2590 goto attr_msg_full;
2591 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2592 goto attr_msg_full;
2593
2594 prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2595 if (!prop)
2596 goto attr_msg_full;
2597 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
2598 goto prop_msg_full;
2599 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2600 goto prop_msg_full;
2601 if (bc_mode & BCLINK_MODE_SEL)
2602 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2603 bc_ratio))
2604 goto prop_msg_full;
2605 nla_nest_end(msg->skb, prop);
2606
2607 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2608 if (err)
2609 goto attr_msg_full;
2610
2611 tipc_bcast_unlock(net);
2612 nla_nest_end(msg->skb, attrs);
2613 genlmsg_end(msg->skb, hdr);
2614
2615 return 0;
2616
2617prop_msg_full:
2618 nla_nest_cancel(msg->skb, prop);
2619attr_msg_full:
2620 nla_nest_cancel(msg->skb, attrs);
2621msg_full:
2622 tipc_bcast_unlock(net);
2623 genlmsg_cancel(msg->skb, hdr);
2624
2625 return -EMSGSIZE;
2626}
2627
2628void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2629 struct sk_buff_head *xmitq)
2630{
2631 l->tolerance = tol;
2632 if (l->bc_rcvlink)
2633 l->bc_rcvlink->tolerance = tol;
2634 if (link_is_up(l))
2635 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2636}
2637
2638void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2639 struct sk_buff_head *xmitq)
2640{
2641 l->priority = prio;
2642 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2643}
2644
2645void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2646{
2647 l->abort_limit = limit;
2648}
2649
2650char *tipc_link_name_ext(struct tipc_link *l, char *buf)
2651{
2652 if (!l)
2653 scnprintf(buf, TIPC_MAX_LINK_NAME, "null");
2654 else if (link_is_bc_sndlink(l))
2655 scnprintf(buf, TIPC_MAX_LINK_NAME, "broadcast-sender");
2656 else if (link_is_bc_rcvlink(l))
2657 scnprintf(buf, TIPC_MAX_LINK_NAME,
2658 "broadcast-receiver, peer %x", l->addr);
2659 else
2660 memcpy(buf, l->name, TIPC_MAX_LINK_NAME);
2661
2662 return buf;
2663}
2664
2665/**
2666 * tipc_link_dump - dump TIPC link data
2667 * @l: tipc link to be dumped
2668 * @dqueues: bitmask to decide if any link queue to be dumped?
2669 * - TIPC_DUMP_NONE: don't dump link queues
2670 * - TIPC_DUMP_TRANSMQ: dump link transmq queue
2671 * - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2672 * - TIPC_DUMP_DEFERDQ: dump link deferd queue
2673 * - TIPC_DUMP_INPUTQ: dump link input queue
2674 * - TIPC_DUMP_WAKEUP: dump link wakeup queue
2675 * - TIPC_DUMP_ALL: dump all the link queues above
2676 * @buf: returned buffer of dump data in format
2677 */
2678int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2679{
2680 int i = 0;
2681 size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2682 struct sk_buff_head *list;
2683 struct sk_buff *hskb, *tskb;
2684 u32 len;
2685
2686 if (!l) {
2687 i += scnprintf(buf, sz, "link data: (null)\n");
2688 return i;
2689 }
2690
2691 i += scnprintf(buf, sz, "link data: %x", l->addr);
2692 i += scnprintf(buf + i, sz - i, " %x", l->state);
2693 i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2694 i += scnprintf(buf + i, sz - i, " %u", l->session);
2695 i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2696 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2697 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2698 i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2699 i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2700 i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2701 i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2702 i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2703 i += scnprintf(buf + i, sz - i, " %u", 0);
2704 i += scnprintf(buf + i, sz - i, " %u", 0);
2705 i += scnprintf(buf + i, sz - i, " %u", l->acked);
2706
2707 list = &l->transmq;
2708 len = skb_queue_len(list);
2709 hskb = skb_peek(list);
2710 tskb = skb_peek_tail(list);
2711 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2712 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2713 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2714
2715 list = &l->deferdq;
2716 len = skb_queue_len(list);
2717 hskb = skb_peek(list);
2718 tskb = skb_peek_tail(list);
2719 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2720 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2721 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2722
2723 list = &l->backlogq;
2724 len = skb_queue_len(list);
2725 hskb = skb_peek(list);
2726 tskb = skb_peek_tail(list);
2727 i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2728 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2729 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2730
2731 list = l->inputq;
2732 len = skb_queue_len(list);
2733 hskb = skb_peek(list);
2734 tskb = skb_peek_tail(list);
2735 i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2736 (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2737 (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2738
2739 if (dqueues & TIPC_DUMP_TRANSMQ) {
2740 i += scnprintf(buf + i, sz - i, "transmq: ");
2741 i += tipc_list_dump(&l->transmq, false, buf + i);
2742 }
2743 if (dqueues & TIPC_DUMP_BACKLOGQ) {
2744 i += scnprintf(buf + i, sz - i,
2745 "backlogq: <%u %u %u %u %u>, ",
2746 l->backlog[TIPC_LOW_IMPORTANCE].len,
2747 l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2748 l->backlog[TIPC_HIGH_IMPORTANCE].len,
2749 l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2750 l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2751 i += tipc_list_dump(&l->backlogq, false, buf + i);
2752 }
2753 if (dqueues & TIPC_DUMP_DEFERDQ) {
2754 i += scnprintf(buf + i, sz - i, "deferdq: ");
2755 i += tipc_list_dump(&l->deferdq, false, buf + i);
2756 }
2757 if (dqueues & TIPC_DUMP_INPUTQ) {
2758 i += scnprintf(buf + i, sz - i, "inputq: ");
2759 i += tipc_list_dump(l->inputq, false, buf + i);
2760 }
2761 if (dqueues & TIPC_DUMP_WAKEUP) {
2762 i += scnprintf(buf + i, sz - i, "wakeup: ");
2763 i += tipc_list_dump(&l->wakeupq, false, buf + i);
2764 }
2765
2766 return i;
2767}