2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
42 #include "name_distr.h"
47 #include <linux/pkt_sched.h>
50 u32 sent_info; /* used in counting # sent packets */
51 u32 recv_info; /* used in counting # recv'd packets */
68 u32 link_congs; /* # port sends blocked by congestion */
71 u32 max_queue_sz; /* send queue size high water mark */
72 u32 accu_queue_sz; /* used for send queue size profiling */
73 u32 queue_sz_counts; /* used for send queue size profiling */
74 u32 msg_length_counts; /* used for message length profiling */
75 u32 msg_lengths_total; /* used for message length profiling */
76 u32 msg_length_profile[7]; /* used for msg. length profiling */
80 * struct tipc_link - TIPC link data structure
81 * @addr: network address of link's peer node
82 * @name: link name character string
83 * @media_addr: media address to use when sending messages over link
85 * @net: pointer to namespace struct
86 * @refcnt: reference counter for permanent references (owner node & timer)
87 * @peer_session: link session # being used by peer end of link
88 * @peer_bearer_id: bearer id used by link's peer endpoint
89 * @bearer_id: local bearer id used by link
90 * @tolerance: minimum link continuity loss needed to reset link [in ms]
91 * @abort_limit: # of unacknowledged continuity probes needed to reset link
92 * @state: current state of link FSM
93 * @peer_caps: bitmap describing capabilities of peer node
94 * @silent_intv_cnt: # of timer intervals without any reception from peer
95 * @proto_msg: template for control messages generated by link
96 * @pmsg: convenience pointer to "proto_msg" field
97 * @priority: current link priority
98 * @net_plane: current link network plane ('A' through 'H')
99 * @mon_state: cookie with information needed by link monitor
100 * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
101 * @exp_msg_count: # of tunnelled messages expected during link changeover
102 * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
103 * @mtu: current maximum packet size for this link
104 * @advertised_mtu: advertised own mtu when link is being established
105 * @transmitq: queue for sent, non-acked messages
106 * @backlogq: queue for messages waiting to be sent
107 * @snt_nxt: next sequence number to use for outbound messages
108 * @last_retransmitted: sequence number of most recently retransmitted message
109 * @stale_count: # of identical retransmit requests made by peer
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
126 char name[TIPC_MAX_LINK_NAME];
129 /* Management and link supervision data */
140 char if_name[TIPC_MAX_IF_NAME];
143 struct tipc_mon_state mon_state;
148 struct sk_buff *failover_reasm_skb;
150 /* Max packet negotiation */
155 struct sk_buff_head transmq;
156 struct sk_buff_head backlogq;
169 struct sk_buff_head deferdq;
170 struct sk_buff_head *inputq;
171 struct sk_buff_head *namedq;
173 /* Congestion handling */
174 struct sk_buff_head wakeupq;
176 /* Fragmentation/reassembly */
177 struct sk_buff *reasm_buf;
182 struct tipc_link *bc_rcvlink;
183 struct tipc_link *bc_sndlink;
188 struct tipc_stats stats;
192 * Error message prefixes
194 static const char *link_co_err = "Link tunneling error, ";
195 static const char *link_rst_msg = "Resetting link ";
197 /* Send states for broadcast NACKs
200 BC_NACK_SND_CONDITIONAL,
201 BC_NACK_SND_UNCONDITIONAL,
202 BC_NACK_SND_SUPPRESS,
206 * Interval between NACKs when packets arrive out of order
208 #define TIPC_NACK_INTV (TIPC_MIN_LINK_WIN * 2)
210 /* Wildcard value for link session numbers. When it is known that
211 * peer endpoint is down, any session number must be accepted.
213 #define ANY_SESSION 0x10000
218 LINK_ESTABLISHED = 0xe,
219 LINK_ESTABLISHING = 0xe << 4,
220 LINK_RESET = 0x1 << 8,
221 LINK_RESETTING = 0x2 << 12,
222 LINK_PEER_RESET = 0xd << 16,
223 LINK_FAILINGOVER = 0xf << 20,
224 LINK_SYNCHING = 0xc << 24
227 /* Link FSM state checking routines
229 static int link_is_up(struct tipc_link *l)
231 return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
234 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
235 struct sk_buff_head *xmitq);
236 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
237 u16 rcvgap, int tolerance, int priority,
238 struct sk_buff_head *xmitq);
239 static void link_print(struct tipc_link *l, const char *str);
240 static void tipc_link_build_nack_msg(struct tipc_link *l,
241 struct sk_buff_head *xmitq);
242 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
243 struct sk_buff_head *xmitq);
244 static bool tipc_link_release_pkts(struct tipc_link *l, u16 to);
247 * Simple non-static link routines (i.e. referenced outside this file)
249 bool tipc_link_is_up(struct tipc_link *l)
251 return link_is_up(l);
254 bool tipc_link_peer_is_down(struct tipc_link *l)
256 return l->state == LINK_PEER_RESET;
259 bool tipc_link_is_reset(struct tipc_link *l)
261 return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
264 bool tipc_link_is_establishing(struct tipc_link *l)
266 return l->state == LINK_ESTABLISHING;
269 bool tipc_link_is_synching(struct tipc_link *l)
271 return l->state == LINK_SYNCHING;
274 bool tipc_link_is_failingover(struct tipc_link *l)
276 return l->state == LINK_FAILINGOVER;
279 bool tipc_link_is_blocked(struct tipc_link *l)
281 return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
284 static bool link_is_bc_sndlink(struct tipc_link *l)
286 return !l->bc_sndlink;
289 static bool link_is_bc_rcvlink(struct tipc_link *l)
291 return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
294 int tipc_link_is_active(struct tipc_link *l)
299 void tipc_link_set_active(struct tipc_link *l, bool active)
304 u32 tipc_link_id(struct tipc_link *l)
306 return l->peer_bearer_id << 16 | l->bearer_id;
309 int tipc_link_window(struct tipc_link *l)
314 int tipc_link_prio(struct tipc_link *l)
319 unsigned long tipc_link_tolerance(struct tipc_link *l)
324 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
329 char tipc_link_plane(struct tipc_link *l)
334 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
335 struct tipc_link *uc_l,
336 struct sk_buff_head *xmitq)
338 struct tipc_link *rcv_l = uc_l->bc_rcvlink;
341 rcv_l->acked = snd_l->snd_nxt - 1;
342 snd_l->state = LINK_ESTABLISHED;
343 tipc_link_build_bc_init_msg(uc_l, xmitq);
346 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
347 struct tipc_link *rcv_l,
348 struct sk_buff_head *xmitq)
350 u16 ack = snd_l->snd_nxt - 1;
353 rcv_l->bc_peer_is_up = true;
354 rcv_l->state = LINK_ESTABLISHED;
355 tipc_link_bc_ack_rcv(rcv_l, ack, xmitq);
356 tipc_link_reset(rcv_l);
357 rcv_l->state = LINK_RESET;
358 if (!snd_l->ackers) {
359 tipc_link_reset(snd_l);
360 snd_l->state = LINK_RESET;
361 __skb_queue_purge(xmitq);
365 int tipc_link_bc_peers(struct tipc_link *l)
370 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
375 int tipc_link_mtu(struct tipc_link *l)
380 u16 tipc_link_rcv_nxt(struct tipc_link *l)
385 u16 tipc_link_acked(struct tipc_link *l)
390 char *tipc_link_name(struct tipc_link *l)
396 * tipc_link_create - create a new link
397 * @n: pointer to associated node
398 * @if_name: associated interface name
399 * @bearer_id: id (index) of associated bearer
400 * @tolerance: link tolerance to be used by link
401 * @net_plane: network plane (A,B,c..) this link belongs to
402 * @mtu: mtu to be advertised by link
403 * @priority: priority to be used by link
404 * @window: send window to be used by link
405 * @session: session to be used by link
406 * @ownnode: identity of own node
407 * @peer: node id of peer node
408 * @peer_caps: bitmap describing peer node capabilities
409 * @bc_sndlink: the namespace global link used for broadcast sending
410 * @bc_rcvlink: the peer specific link used for broadcast reception
411 * @inputq: queue to put messages ready for delivery
412 * @namedq: queue to put binding table update messages ready for delivery
413 * @link: return value, pointer to put the created link
415 * Returns true if link was created, otherwise false
417 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
418 int tolerance, char net_plane, u32 mtu, int priority,
419 int window, u32 session, u32 ownnode, u32 peer,
421 struct tipc_link *bc_sndlink,
422 struct tipc_link *bc_rcvlink,
423 struct sk_buff_head *inputq,
424 struct sk_buff_head *namedq,
425 struct tipc_link **link)
429 l = kzalloc(sizeof(*l), GFP_ATOMIC);
433 l->session = session;
435 /* Note: peer i/f name is completed by reset/activate message */
436 sprintf(l->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
437 tipc_zone(ownnode), tipc_cluster(ownnode), tipc_node(ownnode),
438 if_name, tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
439 strcpy(l->if_name, if_name);
441 l->peer_caps = peer_caps;
443 l->peer_session = ANY_SESSION;
444 l->bearer_id = bearer_id;
445 l->tolerance = tolerance;
446 l->net_plane = net_plane;
447 l->advertised_mtu = mtu;
449 l->priority = priority;
450 tipc_link_set_queue_limits(l, window);
452 l->bc_sndlink = bc_sndlink;
453 l->bc_rcvlink = bc_rcvlink;
456 l->state = LINK_RESETTING;
457 __skb_queue_head_init(&l->transmq);
458 __skb_queue_head_init(&l->backlogq);
459 __skb_queue_head_init(&l->deferdq);
460 skb_queue_head_init(&l->wakeupq);
461 skb_queue_head_init(l->inputq);
466 * tipc_link_bc_create - create new link to be used for broadcast
467 * @n: pointer to associated node
468 * @mtu: mtu to be used
469 * @window: send window to be used
470 * @inputq: queue to put messages ready for delivery
471 * @namedq: queue to put binding table update messages ready for delivery
472 * @link: return value, pointer to put the created link
474 * Returns true if link was created, otherwise false
476 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer,
477 int mtu, int window, u16 peer_caps,
478 struct sk_buff_head *inputq,
479 struct sk_buff_head *namedq,
480 struct tipc_link *bc_sndlink,
481 struct tipc_link **link)
485 if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, window,
486 0, ownnode, peer, peer_caps, bc_sndlink,
487 NULL, inputq, namedq, link))
491 strcpy(l->name, tipc_bclink_name);
493 l->state = LINK_RESET;
497 /* Broadcast send link is always up */
498 if (link_is_bc_sndlink(l))
499 l->state = LINK_ESTABLISHED;
505 * tipc_link_fsm_evt - link finite state machine
506 * @l: pointer to link
507 * @evt: state machine event to be processed
509 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
516 case LINK_PEER_RESET_EVT:
517 l->state = LINK_PEER_RESET;
520 l->state = LINK_RESET;
522 case LINK_FAILURE_EVT:
523 case LINK_FAILOVER_BEGIN_EVT:
524 case LINK_ESTABLISH_EVT:
525 case LINK_FAILOVER_END_EVT:
526 case LINK_SYNCH_BEGIN_EVT:
527 case LINK_SYNCH_END_EVT:
534 case LINK_PEER_RESET_EVT:
535 l->state = LINK_ESTABLISHING;
537 case LINK_FAILOVER_BEGIN_EVT:
538 l->state = LINK_FAILINGOVER;
539 case LINK_FAILURE_EVT:
541 case LINK_ESTABLISH_EVT:
542 case LINK_FAILOVER_END_EVT:
544 case LINK_SYNCH_BEGIN_EVT:
545 case LINK_SYNCH_END_EVT:
550 case LINK_PEER_RESET:
553 l->state = LINK_ESTABLISHING;
555 case LINK_PEER_RESET_EVT:
556 case LINK_ESTABLISH_EVT:
557 case LINK_FAILURE_EVT:
559 case LINK_SYNCH_BEGIN_EVT:
560 case LINK_SYNCH_END_EVT:
561 case LINK_FAILOVER_BEGIN_EVT:
562 case LINK_FAILOVER_END_EVT:
567 case LINK_FAILINGOVER:
569 case LINK_FAILOVER_END_EVT:
570 l->state = LINK_RESET;
572 case LINK_PEER_RESET_EVT:
574 case LINK_ESTABLISH_EVT:
575 case LINK_FAILURE_EVT:
577 case LINK_FAILOVER_BEGIN_EVT:
578 case LINK_SYNCH_BEGIN_EVT:
579 case LINK_SYNCH_END_EVT:
584 case LINK_ESTABLISHING:
586 case LINK_ESTABLISH_EVT:
587 l->state = LINK_ESTABLISHED;
589 case LINK_FAILOVER_BEGIN_EVT:
590 l->state = LINK_FAILINGOVER;
593 l->state = LINK_RESET;
595 case LINK_FAILURE_EVT:
596 case LINK_PEER_RESET_EVT:
597 case LINK_SYNCH_BEGIN_EVT:
598 case LINK_FAILOVER_END_EVT:
600 case LINK_SYNCH_END_EVT:
605 case LINK_ESTABLISHED:
607 case LINK_PEER_RESET_EVT:
608 l->state = LINK_PEER_RESET;
609 rc |= TIPC_LINK_DOWN_EVT;
611 case LINK_FAILURE_EVT:
612 l->state = LINK_RESETTING;
613 rc |= TIPC_LINK_DOWN_EVT;
616 l->state = LINK_RESET;
618 case LINK_ESTABLISH_EVT:
619 case LINK_SYNCH_END_EVT:
621 case LINK_SYNCH_BEGIN_EVT:
622 l->state = LINK_SYNCHING;
624 case LINK_FAILOVER_BEGIN_EVT:
625 case LINK_FAILOVER_END_EVT:
632 case LINK_PEER_RESET_EVT:
633 l->state = LINK_PEER_RESET;
634 rc |= TIPC_LINK_DOWN_EVT;
636 case LINK_FAILURE_EVT:
637 l->state = LINK_RESETTING;
638 rc |= TIPC_LINK_DOWN_EVT;
641 l->state = LINK_RESET;
643 case LINK_ESTABLISH_EVT:
644 case LINK_SYNCH_BEGIN_EVT:
646 case LINK_SYNCH_END_EVT:
647 l->state = LINK_ESTABLISHED;
649 case LINK_FAILOVER_BEGIN_EVT:
650 case LINK_FAILOVER_END_EVT:
656 pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
660 pr_err("Illegal FSM event %x in state %x on link %s\n",
661 evt, l->state, l->name);
665 /* link_profile_stats - update statistical profiling of traffic
667 static void link_profile_stats(struct tipc_link *l)
670 struct tipc_msg *msg;
673 /* Update counters used in statistical profiling of send traffic */
674 l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
675 l->stats.queue_sz_counts++;
677 skb = skb_peek(&l->transmq);
681 length = msg_size(msg);
683 if (msg_user(msg) == MSG_FRAGMENTER) {
684 if (msg_type(msg) != FIRST_FRAGMENT)
686 length = msg_size(msg_get_wrapped(msg));
688 l->stats.msg_lengths_total += length;
689 l->stats.msg_length_counts++;
691 l->stats.msg_length_profile[0]++;
692 else if (length <= 256)
693 l->stats.msg_length_profile[1]++;
694 else if (length <= 1024)
695 l->stats.msg_length_profile[2]++;
696 else if (length <= 4096)
697 l->stats.msg_length_profile[3]++;
698 else if (length <= 16384)
699 l->stats.msg_length_profile[4]++;
700 else if (length <= 32768)
701 l->stats.msg_length_profile[5]++;
703 l->stats.msg_length_profile[6]++;
706 /* tipc_link_timeout - perform periodic task as instructed from node timeout
708 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
715 u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
716 u16 bc_acked = l->bc_rcvlink->acked;
717 struct tipc_mon_state *mstate = &l->mon_state;
720 case LINK_ESTABLISHED:
723 link_profile_stats(l);
724 tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
725 if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
726 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
727 state = bc_acked != bc_snt;
728 state |= l->bc_rcvlink->rcv_unacked;
729 state |= l->rcv_unacked;
730 state |= !skb_queue_empty(&l->transmq);
731 state |= !skb_queue_empty(&l->deferdq);
732 probe = mstate->probing;
733 probe |= l->silent_intv_cnt;
734 if (probe || mstate->monitoring)
735 l->silent_intv_cnt++;
738 setup = l->rst_cnt++ <= 4;
739 setup |= !(l->rst_cnt % 16);
742 case LINK_ESTABLISHING:
746 case LINK_PEER_RESET:
748 case LINK_FAILINGOVER:
754 if (state || probe || setup)
755 tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, xmitq);
761 * link_schedule_user - schedule a message sender for wakeup after congestion
762 * @link: congested link
763 * @list: message that was attempted sent
764 * Create pseudo msg to send back to user when congestion abates
765 * Does not consume buffer list
767 static int link_schedule_user(struct tipc_link *link, struct sk_buff_head *list)
769 struct tipc_msg *msg = buf_msg(skb_peek(list));
770 int imp = msg_importance(msg);
771 u32 oport = msg_origport(msg);
772 u32 addr = tipc_own_addr(link->net);
775 /* This really cannot happen... */
776 if (unlikely(imp > TIPC_CRITICAL_IMPORTANCE)) {
777 pr_warn("%s<%s>, send queue full", link_rst_msg, link->name);
780 /* Non-blocking sender: */
781 if (TIPC_SKB_CB(skb_peek(list))->wakeup_pending)
784 /* Create and schedule wakeup pseudo message */
785 skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
786 addr, addr, oport, 0, 0);
789 TIPC_SKB_CB(skb)->chain_sz = skb_queue_len(list);
790 TIPC_SKB_CB(skb)->chain_imp = imp;
791 skb_queue_tail(&link->wakeupq, skb);
792 link->stats.link_congs++;
797 * link_prepare_wakeup - prepare users for wakeup after congestion
798 * @link: congested link
799 * Move a number of waiting users, as permitted by available space in
800 * the send queue, from link wait queue to node wait queue for wakeup
802 void link_prepare_wakeup(struct tipc_link *l)
804 int pnd[TIPC_SYSTEM_IMPORTANCE + 1] = {0,};
806 struct sk_buff *skb, *tmp;
808 skb_queue_walk_safe(&l->wakeupq, skb, tmp) {
809 imp = TIPC_SKB_CB(skb)->chain_imp;
810 lim = l->backlog[imp].limit;
811 pnd[imp] += TIPC_SKB_CB(skb)->chain_sz;
812 if ((pnd[imp] + l->backlog[imp].len) >= lim)
814 skb_unlink(skb, &l->wakeupq);
815 skb_queue_tail(l->inputq, skb);
819 void tipc_link_reset(struct tipc_link *l)
821 l->peer_session = ANY_SESSION;
823 l->mtu = l->advertised_mtu;
824 __skb_queue_purge(&l->transmq);
825 __skb_queue_purge(&l->deferdq);
826 skb_queue_splice_init(&l->wakeupq, l->inputq);
827 __skb_queue_purge(&l->backlogq);
828 l->backlog[TIPC_LOW_IMPORTANCE].len = 0;
829 l->backlog[TIPC_MEDIUM_IMPORTANCE].len = 0;
830 l->backlog[TIPC_HIGH_IMPORTANCE].len = 0;
831 l->backlog[TIPC_CRITICAL_IMPORTANCE].len = 0;
832 l->backlog[TIPC_SYSTEM_IMPORTANCE].len = 0;
833 kfree_skb(l->reasm_buf);
834 kfree_skb(l->failover_reasm_skb);
836 l->failover_reasm_skb = NULL;
841 l->silent_intv_cnt = 0;
843 l->stats.recv_info = 0;
845 l->bc_peer_is_up = false;
846 memset(&l->mon_state, 0, sizeof(l->mon_state));
847 tipc_link_reset_stats(l);
851 * tipc_link_xmit(): enqueue buffer list according to queue situation
853 * @list: chain of buffers containing message
854 * @xmitq: returned list of packets to be sent by caller
856 * Consumes the buffer chain, except when returning -ELINKCONG,
857 * since the caller then may want to make more send attempts.
858 * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
859 * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
861 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
862 struct sk_buff_head *xmitq)
864 struct tipc_msg *hdr = buf_msg(skb_peek(list));
865 unsigned int maxwin = l->window;
866 unsigned int i, imp = msg_importance(hdr);
867 unsigned int mtu = l->mtu;
868 u16 ack = l->rcv_nxt - 1;
869 u16 seqno = l->snd_nxt;
870 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
871 struct sk_buff_head *transmq = &l->transmq;
872 struct sk_buff_head *backlogq = &l->backlogq;
873 struct sk_buff *skb, *_skb, *bskb;
875 /* Match msg importance against this and all higher backlog limits: */
876 if (!skb_queue_empty(backlogq)) {
877 for (i = imp; i <= TIPC_SYSTEM_IMPORTANCE; i++) {
878 if (unlikely(l->backlog[i].len >= l->backlog[i].limit))
879 return link_schedule_user(l, list);
882 if (unlikely(msg_size(hdr) > mtu)) {
883 skb_queue_purge(list);
887 /* Prepare each packet for sending, and add to relevant queue: */
888 while (skb_queue_len(list)) {
889 skb = skb_peek(list);
891 msg_set_seqno(hdr, seqno);
892 msg_set_ack(hdr, ack);
893 msg_set_bcast_ack(hdr, bc_ack);
895 if (likely(skb_queue_len(transmq) < maxwin)) {
896 _skb = skb_clone(skb, GFP_ATOMIC);
898 skb_queue_purge(list);
902 __skb_queue_tail(transmq, skb);
903 __skb_queue_tail(xmitq, _skb);
904 TIPC_SKB_CB(skb)->ackers = l->ackers;
909 if (tipc_msg_bundle(skb_peek_tail(backlogq), hdr, mtu)) {
910 kfree_skb(__skb_dequeue(list));
911 l->stats.sent_bundled++;
914 if (tipc_msg_make_bundle(&bskb, hdr, mtu, l->addr)) {
915 kfree_skb(__skb_dequeue(list));
916 __skb_queue_tail(backlogq, bskb);
917 l->backlog[msg_importance(buf_msg(bskb))].len++;
918 l->stats.sent_bundled++;
919 l->stats.sent_bundles++;
922 l->backlog[imp].len += skb_queue_len(list);
923 skb_queue_splice_tail_init(list, backlogq);
929 void tipc_link_advance_backlog(struct tipc_link *l, struct sk_buff_head *xmitq)
931 struct sk_buff *skb, *_skb;
932 struct tipc_msg *hdr;
933 u16 seqno = l->snd_nxt;
934 u16 ack = l->rcv_nxt - 1;
935 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
937 while (skb_queue_len(&l->transmq) < l->window) {
938 skb = skb_peek(&l->backlogq);
941 _skb = skb_clone(skb, GFP_ATOMIC);
944 __skb_dequeue(&l->backlogq);
946 l->backlog[msg_importance(hdr)].len--;
947 __skb_queue_tail(&l->transmq, skb);
948 __skb_queue_tail(xmitq, _skb);
949 TIPC_SKB_CB(skb)->ackers = l->ackers;
950 msg_set_seqno(hdr, seqno);
951 msg_set_ack(hdr, ack);
952 msg_set_bcast_ack(hdr, bc_ack);
959 static void link_retransmit_failure(struct tipc_link *l, struct sk_buff *skb)
961 struct tipc_msg *hdr = buf_msg(skb);
963 pr_warn("Retransmission failure on link <%s>\n", l->name);
964 link_print(l, "Resetting link ");
965 pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
966 msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
967 pr_info("sqno %u, prev: %x, src: %x\n",
968 msg_seqno(hdr), msg_prevnode(hdr), msg_orignode(hdr));
971 int tipc_link_retrans(struct tipc_link *l, u16 from, u16 to,
972 struct sk_buff_head *xmitq)
974 struct sk_buff *_skb, *skb = skb_peek(&l->transmq);
975 struct tipc_msg *hdr;
976 u16 ack = l->rcv_nxt - 1;
977 u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
982 /* Detect repeated retransmit failures on same packet */
983 if (likely(l->last_retransm != buf_seqno(skb))) {
984 l->last_retransm = buf_seqno(skb);
986 } else if (++l->stale_count > 100) {
987 link_retransmit_failure(l, skb);
988 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
991 /* Move forward to where retransmission should start */
992 skb_queue_walk(&l->transmq, skb) {
993 if (!less(buf_seqno(skb), from))
997 skb_queue_walk_from(&l->transmq, skb) {
998 if (more(buf_seqno(skb), to))
1001 _skb = __pskb_copy(skb, MIN_H_SIZE, GFP_ATOMIC);
1004 hdr = buf_msg(_skb);
1005 msg_set_ack(hdr, ack);
1006 msg_set_bcast_ack(hdr, bc_ack);
1007 _skb->priority = TC_PRIO_CONTROL;
1008 __skb_queue_tail(xmitq, _skb);
1009 l->stats.retransmitted++;
1014 /* tipc_data_input - deliver data and name distr msgs to upper layer
1016 * Consumes buffer if message is of right type
1017 * Node lock must be held
1019 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1020 struct sk_buff_head *inputq)
1022 switch (msg_user(buf_msg(skb))) {
1023 case TIPC_LOW_IMPORTANCE:
1024 case TIPC_MEDIUM_IMPORTANCE:
1025 case TIPC_HIGH_IMPORTANCE:
1026 case TIPC_CRITICAL_IMPORTANCE:
1028 skb_queue_tail(inputq, skb);
1030 case NAME_DISTRIBUTOR:
1031 l->bc_rcvlink->state = LINK_ESTABLISHED;
1032 skb_queue_tail(l->namedq, skb);
1035 case TUNNEL_PROTOCOL:
1036 case MSG_FRAGMENTER:
1037 case BCAST_PROTOCOL:
1040 pr_warn("Dropping received illegal msg type\n");
1046 /* tipc_link_input - process packet that has passed link protocol check
1050 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1051 struct sk_buff_head *inputq)
1053 struct tipc_msg *hdr = buf_msg(skb);
1054 struct sk_buff **reasm_skb = &l->reasm_buf;
1055 struct sk_buff *iskb;
1056 struct sk_buff_head tmpq;
1057 int usr = msg_user(hdr);
1062 if (unlikely(usr == TUNNEL_PROTOCOL)) {
1063 if (msg_type(hdr) == SYNCH_MSG) {
1064 __skb_queue_purge(&l->deferdq);
1067 if (!tipc_msg_extract(skb, &iskb, &ipos))
1072 if (less(msg_seqno(hdr), l->drop_point))
1074 if (tipc_data_input(l, skb, inputq))
1076 usr = msg_user(hdr);
1077 reasm_skb = &l->failover_reasm_skb;
1080 if (usr == MSG_BUNDLER) {
1081 skb_queue_head_init(&tmpq);
1082 l->stats.recv_bundles++;
1083 l->stats.recv_bundled += msg_msgcnt(hdr);
1084 while (tipc_msg_extract(skb, &iskb, &pos))
1085 tipc_data_input(l, iskb, &tmpq);
1086 tipc_skb_queue_splice_tail(&tmpq, inputq);
1088 } else if (usr == MSG_FRAGMENTER) {
1089 l->stats.recv_fragments++;
1090 if (tipc_buf_append(reasm_skb, &skb)) {
1091 l->stats.recv_fragmented++;
1092 tipc_data_input(l, skb, inputq);
1093 } else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1094 pr_warn_ratelimited("Unable to build fragment list\n");
1095 return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1098 } else if (usr == BCAST_PROTOCOL) {
1099 tipc_bcast_lock(l->net);
1100 tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1101 tipc_bcast_unlock(l->net);
1108 static bool tipc_link_release_pkts(struct tipc_link *l, u16 acked)
1110 bool released = false;
1111 struct sk_buff *skb, *tmp;
1113 skb_queue_walk_safe(&l->transmq, skb, tmp) {
1114 if (more(buf_seqno(skb), acked))
1116 __skb_unlink(skb, &l->transmq);
1123 /* tipc_link_build_state_msg: prepare link state message for transmission
1125 * Note that sending of broadcast ack is coordinated among nodes, to reduce
1126 * risk of ack storms towards the sender
1128 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1133 /* Broadcast ACK must be sent via a unicast link => defer to caller */
1134 if (link_is_bc_rcvlink(l)) {
1135 if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1138 return TIPC_LINK_SND_BC_ACK;
1143 l->stats.sent_acks++;
1144 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1148 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1150 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1152 int mtyp = RESET_MSG;
1153 struct sk_buff *skb;
1155 if (l->state == LINK_ESTABLISHING)
1156 mtyp = ACTIVATE_MSG;
1158 tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, xmitq);
1160 /* Inform peer that this endpoint is going down if applicable */
1161 skb = skb_peek_tail(xmitq);
1162 if (skb && (l->state == LINK_RESET))
1163 msg_set_peer_stopping(buf_msg(skb), 1);
1166 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1168 static void tipc_link_build_nack_msg(struct tipc_link *l,
1169 struct sk_buff_head *xmitq)
1171 u32 def_cnt = ++l->stats.deferred_recv;
1173 if (link_is_bc_rcvlink(l))
1176 if ((skb_queue_len(&l->deferdq) == 1) || !(def_cnt % TIPC_NACK_INTV))
1177 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, xmitq);
1180 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1181 * @l: the link that should handle the message
1183 * @xmitq: queue to place packets to be sent after this call
1185 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1186 struct sk_buff_head *xmitq)
1188 struct sk_buff_head *defq = &l->deferdq;
1189 struct tipc_msg *hdr;
1190 u16 seqno, rcv_nxt, win_lim;
1195 seqno = msg_seqno(hdr);
1196 rcv_nxt = l->rcv_nxt;
1197 win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1199 /* Verify and update link state */
1200 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1201 return tipc_link_proto_rcv(l, skb, xmitq);
1203 if (unlikely(!link_is_up(l))) {
1204 if (l->state == LINK_ESTABLISHING)
1205 rc = TIPC_LINK_UP_EVT;
1209 /* Don't send probe at next timeout expiration */
1210 l->silent_intv_cnt = 0;
1212 /* Drop if outside receive window */
1213 if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1214 l->stats.duplicates++;
1218 /* Forward queues and wake up waiting users */
1219 if (likely(tipc_link_release_pkts(l, msg_ack(hdr)))) {
1220 tipc_link_advance_backlog(l, xmitq);
1221 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1222 link_prepare_wakeup(l);
1225 /* Defer delivery if sequence gap */
1226 if (unlikely(seqno != rcv_nxt)) {
1227 __tipc_skb_queue_sorted(defq, seqno, skb);
1228 tipc_link_build_nack_msg(l, xmitq);
1232 /* Deliver packet */
1234 l->stats.recv_info++;
1235 if (!tipc_data_input(l, skb, l->inputq))
1236 rc |= tipc_link_input(l, skb, l->inputq);
1237 if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1238 rc |= tipc_link_build_state_msg(l, xmitq);
1239 if (unlikely(rc & ~TIPC_LINK_SND_BC_ACK))
1241 } while ((skb = __skb_dequeue(defq)));
1249 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1250 u16 rcvgap, int tolerance, int priority,
1251 struct sk_buff_head *xmitq)
1253 struct sk_buff *skb;
1254 struct tipc_msg *hdr;
1255 struct sk_buff_head *dfq = &l->deferdq;
1256 bool node_up = link_is_up(l->bc_rcvlink);
1257 struct tipc_mon_state *mstate = &l->mon_state;
1261 /* Don't send protocol message during reset or link failover */
1262 if (tipc_link_is_blocked(l))
1265 if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1268 if (!skb_queue_empty(dfq))
1269 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1271 skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1272 tipc_max_domain_size, l->addr,
1273 tipc_own_addr(l->net), 0, 0, 0);
1278 data = msg_data(hdr);
1279 msg_set_session(hdr, l->session);
1280 msg_set_bearer_id(hdr, l->bearer_id);
1281 msg_set_net_plane(hdr, l->net_plane);
1282 msg_set_next_sent(hdr, l->snd_nxt);
1283 msg_set_ack(hdr, l->rcv_nxt - 1);
1284 msg_set_bcast_ack(hdr, l->bc_rcvlink->rcv_nxt - 1);
1285 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1286 msg_set_link_tolerance(hdr, tolerance);
1287 msg_set_linkprio(hdr, priority);
1288 msg_set_redundant_link(hdr, node_up);
1289 msg_set_seq_gap(hdr, 0);
1290 msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1292 if (mtyp == STATE_MSG) {
1293 msg_set_seq_gap(hdr, rcvgap);
1294 msg_set_probe(hdr, probe);
1295 tipc_mon_prep(l->net, data, &dlen, mstate, l->bearer_id);
1296 msg_set_size(hdr, INT_H_SIZE + dlen);
1297 skb_trim(skb, INT_H_SIZE + dlen);
1298 l->stats.sent_states++;
1301 /* RESET_MSG or ACTIVATE_MSG */
1302 msg_set_max_pkt(hdr, l->advertised_mtu);
1303 strcpy(data, l->if_name);
1304 msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1305 skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1308 l->stats.sent_probes++;
1310 l->stats.sent_nacks++;
1311 skb->priority = TC_PRIO_CONTROL;
1312 __skb_queue_tail(xmitq, skb);
1315 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1316 * with contents of the link's transmit and backlog queues.
1318 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1319 int mtyp, struct sk_buff_head *xmitq)
1321 struct sk_buff *skb, *tnlskb;
1322 struct tipc_msg *hdr, tnlhdr;
1323 struct sk_buff_head *queue = &l->transmq;
1324 struct sk_buff_head tmpxq, tnlq;
1325 u16 pktlen, pktcnt, seqno = l->snd_nxt;
1330 skb_queue_head_init(&tnlq);
1331 skb_queue_head_init(&tmpxq);
1333 /* At least one packet required for safe algorithm => add dummy */
1334 skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1335 BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1336 0, 0, TIPC_ERR_NO_PORT);
1338 pr_warn("%sunable to create tunnel packet\n", link_co_err);
1341 skb_queue_tail(&tnlq, skb);
1342 tipc_link_xmit(l, &tnlq, &tmpxq);
1343 __skb_queue_purge(&tmpxq);
1345 /* Initialize reusable tunnel packet header */
1346 tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1347 mtyp, INT_H_SIZE, l->addr);
1348 pktcnt = skb_queue_len(&l->transmq) + skb_queue_len(&l->backlogq);
1349 msg_set_msgcnt(&tnlhdr, pktcnt);
1350 msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1352 /* Wrap each packet into a tunnel packet */
1353 skb_queue_walk(queue, skb) {
1355 if (queue == &l->backlogq)
1356 msg_set_seqno(hdr, seqno++);
1357 pktlen = msg_size(hdr);
1358 msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
1359 tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE);
1361 pr_warn("%sunable to send packet\n", link_co_err);
1364 skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
1365 skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
1366 __skb_queue_tail(&tnlq, tnlskb);
1368 if (queue != &l->backlogq) {
1369 queue = &l->backlogq;
1373 tipc_link_xmit(tnl, &tnlq, xmitq);
1375 if (mtyp == FAILOVER_MSG) {
1376 tnl->drop_point = l->rcv_nxt;
1377 tnl->failover_reasm_skb = l->reasm_buf;
1378 l->reasm_buf = NULL;
1382 /* tipc_link_proto_rcv(): receive link level protocol message :
1383 * Note that network plane id propagates through the network, and may
1384 * change at any time. The node with lowest numerical id determines
1387 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
1388 struct sk_buff_head *xmitq)
1390 struct tipc_msg *hdr = buf_msg(skb);
1392 u16 ack = msg_ack(hdr);
1393 u16 gap = msg_seq_gap(hdr);
1394 u16 peers_snd_nxt = msg_next_sent(hdr);
1395 u16 peers_tol = msg_link_tolerance(hdr);
1396 u16 peers_prio = msg_linkprio(hdr);
1397 u16 rcv_nxt = l->rcv_nxt;
1398 u16 dlen = msg_data_sz(hdr);
1399 int mtyp = msg_type(hdr);
1404 if (tipc_link_is_blocked(l) || !xmitq)
1407 if (tipc_own_addr(l->net) > msg_prevnode(hdr))
1408 l->net_plane = msg_net_plane(hdr);
1412 data = msg_data(hdr);
1417 /* Ignore duplicate RESET with old session number */
1418 if ((less_eq(msg_session(hdr), l->peer_session)) &&
1419 (l->peer_session != ANY_SESSION))
1425 /* Complete own link name with peer's interface name */
1426 if_name = strrchr(l->name, ':') + 1;
1427 if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
1429 if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
1431 strncpy(if_name, data, TIPC_MAX_IF_NAME);
1433 /* Update own tolerance if peer indicates a non-zero value */
1434 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1435 l->tolerance = peers_tol;
1437 /* Update own priority if peer's priority is higher */
1438 if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
1439 l->priority = peers_prio;
1441 /* ACTIVATE_MSG serves as PEER_RESET if link is already down */
1442 if (msg_peer_stopping(hdr))
1443 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1444 else if ((mtyp == RESET_MSG) || !link_is_up(l))
1445 rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
1447 /* ACTIVATE_MSG takes up link if it was already locally reset */
1448 if ((mtyp == ACTIVATE_MSG) && (l->state == LINK_ESTABLISHING))
1449 rc = TIPC_LINK_UP_EVT;
1451 l->peer_session = msg_session(hdr);
1452 l->peer_bearer_id = msg_bearer_id(hdr);
1453 if (l->mtu > msg_max_pkt(hdr))
1454 l->mtu = msg_max_pkt(hdr);
1459 /* Update own tolerance if peer indicates a non-zero value */
1460 if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL))
1461 l->tolerance = peers_tol;
1463 if (peers_prio && in_range(peers_prio, TIPC_MIN_LINK_PRI,
1464 TIPC_MAX_LINK_PRI)) {
1465 l->priority = peers_prio;
1466 rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1469 l->silent_intv_cnt = 0;
1470 l->stats.recv_states++;
1472 l->stats.recv_probes++;
1474 if (!link_is_up(l)) {
1475 if (l->state == LINK_ESTABLISHING)
1476 rc = TIPC_LINK_UP_EVT;
1479 tipc_mon_rcv(l->net, data, dlen, l->addr,
1480 &l->mon_state, l->bearer_id);
1482 /* Send NACK if peer has sent pkts we haven't received yet */
1483 if (more(peers_snd_nxt, rcv_nxt) && !tipc_link_is_synching(l))
1484 rcvgap = peers_snd_nxt - l->rcv_nxt;
1485 if (rcvgap || (msg_probe(hdr)))
1486 tipc_link_build_proto_msg(l, STATE_MSG, 0, rcvgap,
1488 tipc_link_release_pkts(l, ack);
1490 /* If NACK, retransmit will now start at right position */
1492 rc = tipc_link_retrans(l, ack + 1, ack + gap, xmitq);
1493 l->stats.recv_nacks++;
1496 tipc_link_advance_backlog(l, xmitq);
1497 if (unlikely(!skb_queue_empty(&l->wakeupq)))
1498 link_prepare_wakeup(l);
1505 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
1507 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
1509 struct sk_buff_head *xmitq)
1511 struct sk_buff *skb;
1512 struct tipc_msg *hdr;
1513 struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
1514 u16 ack = l->rcv_nxt - 1;
1515 u16 gap_to = peers_snd_nxt - 1;
1517 skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
1518 0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
1522 msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1523 msg_set_bcast_ack(hdr, ack);
1524 msg_set_bcgap_after(hdr, ack);
1526 gap_to = buf_seqno(dfrd_skb) - 1;
1527 msg_set_bcgap_to(hdr, gap_to);
1528 msg_set_non_seq(hdr, bcast);
1529 __skb_queue_tail(xmitq, skb);
1533 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
1535 * Give a newly added peer node the sequence number where it should
1536 * start receiving and acking broadcast packets.
1538 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
1539 struct sk_buff_head *xmitq)
1541 struct sk_buff_head list;
1543 __skb_queue_head_init(&list);
1544 if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
1546 tipc_link_xmit(l, &list, xmitq);
1549 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
1551 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
1553 int mtyp = msg_type(hdr);
1554 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1559 if (msg_user(hdr) == BCAST_PROTOCOL) {
1560 l->rcv_nxt = peers_snd_nxt;
1561 l->state = LINK_ESTABLISHED;
1565 if (l->peer_caps & TIPC_BCAST_SYNCH)
1568 if (msg_peer_node_is_up(hdr))
1571 /* Compatibility: accept older, less safe initial synch data */
1572 if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
1573 l->rcv_nxt = peers_snd_nxt;
1576 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
1578 void tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
1579 struct sk_buff_head *xmitq)
1581 u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
1586 if (!msg_peer_node_is_up(hdr))
1589 /* Open when peer ackowledges our bcast init msg (pkt #1) */
1591 l->bc_peer_is_up = true;
1593 if (!l->bc_peer_is_up)
1596 /* Ignore if peers_snd_nxt goes beyond receive window */
1597 if (more(peers_snd_nxt, l->rcv_nxt + l->window))
1600 if (!more(peers_snd_nxt, l->rcv_nxt)) {
1601 l->nack_state = BC_NACK_SND_CONDITIONAL;
1605 /* Don't NACK if one was recently sent or peeked */
1606 if (l->nack_state == BC_NACK_SND_SUPPRESS) {
1607 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1611 /* Conditionally delay NACK sending until next synch rcv */
1612 if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
1613 l->nack_state = BC_NACK_SND_UNCONDITIONAL;
1614 if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
1618 /* Send NACK now but suppress next one */
1619 tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
1620 l->nack_state = BC_NACK_SND_SUPPRESS;
1623 void tipc_link_bc_ack_rcv(struct tipc_link *l, u16 acked,
1624 struct sk_buff_head *xmitq)
1626 struct sk_buff *skb, *tmp;
1627 struct tipc_link *snd_l = l->bc_sndlink;
1629 if (!link_is_up(l) || !l->bc_peer_is_up)
1632 if (!more(acked, l->acked))
1635 /* Skip over packets peer has already acked */
1636 skb_queue_walk(&snd_l->transmq, skb) {
1637 if (more(buf_seqno(skb), l->acked))
1641 /* Update/release the packets peer is acking now */
1642 skb_queue_walk_from_safe(&snd_l->transmq, skb, tmp) {
1643 if (more(buf_seqno(skb), acked))
1645 if (!--TIPC_SKB_CB(skb)->ackers) {
1646 __skb_unlink(skb, &snd_l->transmq);
1651 tipc_link_advance_backlog(snd_l, xmitq);
1652 if (unlikely(!skb_queue_empty(&snd_l->wakeupq)))
1653 link_prepare_wakeup(snd_l);
1656 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
1658 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
1659 struct sk_buff_head *xmitq)
1661 struct tipc_msg *hdr = buf_msg(skb);
1662 u32 dnode = msg_destnode(hdr);
1663 int mtyp = msg_type(hdr);
1664 u16 acked = msg_bcast_ack(hdr);
1665 u16 from = acked + 1;
1666 u16 to = msg_bcgap_to(hdr);
1667 u16 peers_snd_nxt = to + 1;
1672 if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
1675 if (mtyp != STATE_MSG)
1678 if (dnode == tipc_own_addr(l->net)) {
1679 tipc_link_bc_ack_rcv(l, acked, xmitq);
1680 rc = tipc_link_retrans(l->bc_sndlink, from, to, xmitq);
1681 l->stats.recv_nacks++;
1685 /* Msg for other node => suppress own NACK at next sync if applicable */
1686 if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
1687 l->nack_state = BC_NACK_SND_SUPPRESS;
1692 void tipc_link_set_queue_limits(struct tipc_link *l, u32 win)
1694 int max_bulk = TIPC_MAX_PUBLICATIONS / (l->mtu / ITEM_SIZE);
1697 l->backlog[TIPC_LOW_IMPORTANCE].limit = max_t(u16, 50, win);
1698 l->backlog[TIPC_MEDIUM_IMPORTANCE].limit = max_t(u16, 100, win * 2);
1699 l->backlog[TIPC_HIGH_IMPORTANCE].limit = max_t(u16, 150, win * 3);
1700 l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = max_t(u16, 200, win * 4);
1701 l->backlog[TIPC_SYSTEM_IMPORTANCE].limit = max_bulk;
1705 * link_reset_stats - reset link statistics
1706 * @l: pointer to link
1708 void tipc_link_reset_stats(struct tipc_link *l)
1710 memset(&l->stats, 0, sizeof(l->stats));
1711 if (!link_is_bc_sndlink(l)) {
1712 l->stats.sent_info = l->snd_nxt;
1713 l->stats.recv_info = l->rcv_nxt;
1717 static void link_print(struct tipc_link *l, const char *str)
1719 struct sk_buff *hskb = skb_peek(&l->transmq);
1720 u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
1721 u16 tail = l->snd_nxt - 1;
1723 pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
1724 pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
1725 skb_queue_len(&l->transmq), head, tail,
1726 skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
1729 /* Parse and validate nested (link) properties valid for media, bearer and link
1731 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
1735 err = nla_parse_nested(props, TIPC_NLA_PROP_MAX, prop,
1736 tipc_nl_prop_policy);
1740 if (props[TIPC_NLA_PROP_PRIO]) {
1743 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1744 if (prio > TIPC_MAX_LINK_PRI)
1748 if (props[TIPC_NLA_PROP_TOL]) {
1751 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1752 if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
1756 if (props[TIPC_NLA_PROP_WIN]) {
1759 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1760 if ((win < TIPC_MIN_LINK_WIN) || (win > TIPC_MAX_LINK_WIN))
1767 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
1770 struct nlattr *stats;
1777 struct nla_map map[] = {
1778 {TIPC_NLA_STATS_RX_INFO, s->recv_info},
1779 {TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
1780 {TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
1781 {TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
1782 {TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
1783 {TIPC_NLA_STATS_TX_INFO, s->sent_info},
1784 {TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
1785 {TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
1786 {TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
1787 {TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
1788 {TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
1789 s->msg_length_counts : 1},
1790 {TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
1791 {TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
1792 {TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
1793 {TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
1794 {TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
1795 {TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
1796 {TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
1797 {TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
1798 {TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
1799 {TIPC_NLA_STATS_RX_STATES, s->recv_states},
1800 {TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
1801 {TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
1802 {TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
1803 {TIPC_NLA_STATS_TX_STATES, s->sent_states},
1804 {TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
1805 {TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
1806 {TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
1807 {TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
1808 {TIPC_NLA_STATS_DUPLICATES, s->duplicates},
1809 {TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
1810 {TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
1811 {TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
1812 (s->accu_queue_sz / s->queue_sz_counts) : 0}
1815 stats = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1819 for (i = 0; i < ARRAY_SIZE(map); i++)
1820 if (nla_put_u32(skb, map[i].key, map[i].val))
1823 nla_nest_end(skb, stats);
1827 nla_nest_cancel(skb, stats);
1832 /* Caller should hold appropriate locks to protect the link */
1833 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
1834 struct tipc_link *link, int nlflags)
1838 struct nlattr *attrs;
1839 struct nlattr *prop;
1840 struct tipc_net *tn = net_generic(net, tipc_net_id);
1842 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1843 nlflags, TIPC_NL_LINK_GET);
1847 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1851 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
1853 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST,
1854 tipc_cluster_mask(tn->own_addr)))
1856 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
1858 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->rcv_nxt))
1860 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->snd_nxt))
1863 if (tipc_link_is_up(link))
1864 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1867 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
1870 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1873 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1875 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
1877 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
1880 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
1882 nla_nest_end(msg->skb, prop);
1884 err = __tipc_nl_add_stats(msg->skb, &link->stats);
1888 nla_nest_end(msg->skb, attrs);
1889 genlmsg_end(msg->skb, hdr);
1894 nla_nest_cancel(msg->skb, prop);
1896 nla_nest_cancel(msg->skb, attrs);
1898 genlmsg_cancel(msg->skb, hdr);
1903 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
1904 struct tipc_stats *stats)
1907 struct nlattr *nest;
1914 struct nla_map map[] = {
1915 {TIPC_NLA_STATS_RX_INFO, stats->recv_info},
1916 {TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
1917 {TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
1918 {TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
1919 {TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
1920 {TIPC_NLA_STATS_TX_INFO, stats->sent_info},
1921 {TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
1922 {TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
1923 {TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
1924 {TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
1925 {TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
1926 {TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
1927 {TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
1928 {TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
1929 {TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
1930 {TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
1931 {TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
1932 {TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
1933 {TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
1934 (stats->accu_queue_sz / stats->queue_sz_counts) : 0}
1937 nest = nla_nest_start(skb, TIPC_NLA_LINK_STATS);
1941 for (i = 0; i < ARRAY_SIZE(map); i++)
1942 if (nla_put_u32(skb, map[i].key, map[i].val))
1945 nla_nest_end(skb, nest);
1949 nla_nest_cancel(skb, nest);
1954 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg)
1958 struct nlattr *attrs;
1959 struct nlattr *prop;
1960 struct tipc_net *tn = net_generic(net, tipc_net_id);
1961 struct tipc_link *bcl = tn->bcl;
1966 tipc_bcast_lock(net);
1968 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1969 NLM_F_MULTI, TIPC_NL_LINK_GET);
1971 tipc_bcast_unlock(net);
1975 attrs = nla_nest_start(msg->skb, TIPC_NLA_LINK);
1979 /* The broadcast link is always up */
1980 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
1983 if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
1985 if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
1987 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, bcl->rcv_nxt))
1989 if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, bcl->snd_nxt))
1992 prop = nla_nest_start(msg->skb, TIPC_NLA_LINK_PROP);
1995 if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->window))
1997 nla_nest_end(msg->skb, prop);
1999 err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2003 tipc_bcast_unlock(net);
2004 nla_nest_end(msg->skb, attrs);
2005 genlmsg_end(msg->skb, hdr);
2010 nla_nest_cancel(msg->skb, prop);
2012 nla_nest_cancel(msg->skb, attrs);
2014 tipc_bcast_unlock(net);
2015 genlmsg_cancel(msg->skb, hdr);
2020 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2021 struct sk_buff_head *xmitq)
2024 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, tol, 0, xmitq);
2027 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2028 struct sk_buff_head *xmitq)
2031 tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, prio, xmitq);
2034 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2036 l->abort_limit = limit;