2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2014, 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.
40 #include "name_distr.h"
44 #include <linux/pkt_sched.h>
47 * Error message prefixes
49 static const char *link_co_err = "Link changeover error, ";
50 static const char *link_rst_msg = "Resetting link ";
51 static const char *link_unk_evt = "Unknown link event ";
54 * Out-of-range value for link session numbers
56 #define INVALID_SESSION 0x10000
61 #define STARTING_EVT 856384768 /* link processing trigger */
62 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
63 #define TIMEOUT_EVT 560817u /* link timer expired */
66 * The following two 'message types' is really just implementation
67 * data conveniently stored in the message header.
68 * They must not be considered part of the protocol
74 * State value stored in 'exp_msg_count'
76 #define START_CHANGEOVER 100000u
78 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
80 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf);
81 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
82 struct sk_buff **buf);
83 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
84 static int tipc_link_iovec_long_xmit(struct tipc_port *sender,
85 struct iovec const *msg_sect,
86 unsigned int len, u32 destnode);
87 static void link_state_event(struct tipc_link *l_ptr, u32 event);
88 static void link_reset_statistics(struct tipc_link *l_ptr);
89 static void link_print(struct tipc_link *l_ptr, const char *str);
90 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf);
91 static void tipc_link_sync_xmit(struct tipc_link *l);
92 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
95 * Simple link routines
97 static unsigned int align(unsigned int i)
102 static void link_init_max_pkt(struct tipc_link *l_ptr)
106 max_pkt = (l_ptr->b_ptr->mtu & ~3);
107 if (max_pkt > MAX_MSG_SIZE)
108 max_pkt = MAX_MSG_SIZE;
110 l_ptr->max_pkt_target = max_pkt;
111 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
112 l_ptr->max_pkt = l_ptr->max_pkt_target;
114 l_ptr->max_pkt = MAX_PKT_DEFAULT;
116 l_ptr->max_pkt_probes = 0;
119 static u32 link_next_sent(struct tipc_link *l_ptr)
122 return buf_seqno(l_ptr->next_out);
123 return mod(l_ptr->next_out_no);
126 static u32 link_last_sent(struct tipc_link *l_ptr)
128 return mod(link_next_sent(l_ptr) - 1);
132 * Simple non-static link routines (i.e. referenced outside this file)
134 int tipc_link_is_up(struct tipc_link *l_ptr)
138 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
141 int tipc_link_is_active(struct tipc_link *l_ptr)
143 return (l_ptr->owner->active_links[0] == l_ptr) ||
144 (l_ptr->owner->active_links[1] == l_ptr);
148 * link_timeout - handle expiration of link timer
149 * @l_ptr: pointer to link
151 static void link_timeout(struct tipc_link *l_ptr)
153 tipc_node_lock(l_ptr->owner);
155 /* update counters used in statistical profiling of send traffic */
156 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
157 l_ptr->stats.queue_sz_counts++;
159 if (l_ptr->first_out) {
160 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
161 u32 length = msg_size(msg);
163 if ((msg_user(msg) == MSG_FRAGMENTER) &&
164 (msg_type(msg) == FIRST_FRAGMENT)) {
165 length = msg_size(msg_get_wrapped(msg));
168 l_ptr->stats.msg_lengths_total += length;
169 l_ptr->stats.msg_length_counts++;
171 l_ptr->stats.msg_length_profile[0]++;
172 else if (length <= 256)
173 l_ptr->stats.msg_length_profile[1]++;
174 else if (length <= 1024)
175 l_ptr->stats.msg_length_profile[2]++;
176 else if (length <= 4096)
177 l_ptr->stats.msg_length_profile[3]++;
178 else if (length <= 16384)
179 l_ptr->stats.msg_length_profile[4]++;
180 else if (length <= 32768)
181 l_ptr->stats.msg_length_profile[5]++;
183 l_ptr->stats.msg_length_profile[6]++;
187 /* do all other link processing performed on a periodic basis */
189 link_state_event(l_ptr, TIMEOUT_EVT);
192 tipc_link_push_queue(l_ptr);
194 tipc_node_unlock(l_ptr->owner);
197 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
199 k_start_timer(&l_ptr->timer, time);
203 * tipc_link_create - create a new link
204 * @n_ptr: pointer to associated node
205 * @b_ptr: pointer to associated bearer
206 * @media_addr: media address to use when sending messages over link
208 * Returns pointer to link.
210 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
211 struct tipc_bearer *b_ptr,
212 const struct tipc_media_addr *media_addr)
214 struct tipc_link *l_ptr;
215 struct tipc_msg *msg;
217 char addr_string[16];
218 u32 peer = n_ptr->addr;
220 if (n_ptr->link_cnt >= 2) {
221 tipc_addr_string_fill(addr_string, n_ptr->addr);
222 pr_err("Attempt to establish third link to %s\n", addr_string);
226 if (n_ptr->links[b_ptr->identity]) {
227 tipc_addr_string_fill(addr_string, n_ptr->addr);
228 pr_err("Attempt to establish second link on <%s> to %s\n",
229 b_ptr->name, addr_string);
233 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
235 pr_warn("Link creation failed, no memory\n");
240 if_name = strchr(b_ptr->name, ':') + 1;
241 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
242 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
243 tipc_node(tipc_own_addr),
245 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
246 /* note: peer i/f name is updated by reset/activate message */
247 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
248 l_ptr->owner = n_ptr;
249 l_ptr->checkpoint = 1;
250 l_ptr->peer_session = INVALID_SESSION;
251 l_ptr->b_ptr = b_ptr;
252 link_set_supervision_props(l_ptr, b_ptr->tolerance);
253 l_ptr->state = RESET_UNKNOWN;
255 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
257 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
258 msg_set_size(msg, sizeof(l_ptr->proto_msg));
259 msg_set_session(msg, (tipc_random & 0xffff));
260 msg_set_bearer_id(msg, b_ptr->identity);
261 strcpy((char *)msg_data(msg), if_name);
263 l_ptr->priority = b_ptr->priority;
264 tipc_link_set_queue_limits(l_ptr, b_ptr->window);
266 link_init_max_pkt(l_ptr);
268 l_ptr->next_out_no = 1;
269 INIT_LIST_HEAD(&l_ptr->waiting_ports);
271 link_reset_statistics(l_ptr);
273 tipc_node_attach_link(n_ptr, l_ptr);
275 k_init_timer(&l_ptr->timer, (Handler)link_timeout,
276 (unsigned long)l_ptr);
278 link_state_event(l_ptr, STARTING_EVT);
283 void tipc_link_delete_list(unsigned int bearer_id, bool shutting_down)
285 struct tipc_link *l_ptr;
286 struct tipc_node *n_ptr;
289 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
290 spin_lock_bh(&n_ptr->lock);
291 l_ptr = n_ptr->links[bearer_id];
293 tipc_link_reset(l_ptr);
294 if (shutting_down || !tipc_node_is_up(n_ptr)) {
295 tipc_node_detach_link(l_ptr->owner, l_ptr);
296 tipc_link_reset_fragments(l_ptr);
297 spin_unlock_bh(&n_ptr->lock);
299 /* Nobody else can access this link now: */
300 del_timer_sync(&l_ptr->timer);
303 /* Detach/delete when failover is finished: */
304 l_ptr->flags |= LINK_STOPPED;
305 spin_unlock_bh(&n_ptr->lock);
306 del_timer_sync(&l_ptr->timer);
310 spin_unlock_bh(&n_ptr->lock);
316 * link_schedule_port - schedule port for deferred sending
317 * @l_ptr: pointer to link
318 * @origport: reference to sending port
319 * @sz: amount of data to be sent
321 * Schedules port for renewed sending of messages after link congestion
324 static int link_schedule_port(struct tipc_link *l_ptr, u32 origport, u32 sz)
326 struct tipc_port *p_ptr;
328 spin_lock_bh(&tipc_port_list_lock);
329 p_ptr = tipc_port_lock(origport);
331 if (!list_empty(&p_ptr->wait_list))
333 p_ptr->congested = 1;
334 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
335 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
336 l_ptr->stats.link_congs++;
338 tipc_port_unlock(p_ptr);
340 spin_unlock_bh(&tipc_port_list_lock);
344 void tipc_link_wakeup_ports(struct tipc_link *l_ptr, int all)
346 struct tipc_port *p_ptr;
347 struct tipc_port *temp_p_ptr;
348 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
354 if (!spin_trylock_bh(&tipc_port_list_lock))
356 if (link_congested(l_ptr))
358 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
362 list_del_init(&p_ptr->wait_list);
363 spin_lock_bh(p_ptr->lock);
364 p_ptr->congested = 0;
365 tipc_port_wakeup(p_ptr);
366 win -= p_ptr->waiting_pkts;
367 spin_unlock_bh(p_ptr->lock);
371 spin_unlock_bh(&tipc_port_list_lock);
375 * link_release_outqueue - purge link's outbound message queue
376 * @l_ptr: pointer to link
378 static void link_release_outqueue(struct tipc_link *l_ptr)
380 kfree_skb_list(l_ptr->first_out);
381 l_ptr->first_out = NULL;
382 l_ptr->out_queue_size = 0;
386 * tipc_link_reset_fragments - purge link's inbound message fragments queue
387 * @l_ptr: pointer to link
389 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
391 kfree_skb(l_ptr->reasm_head);
392 l_ptr->reasm_head = NULL;
393 l_ptr->reasm_tail = NULL;
397 * tipc_link_purge_queues - purge all pkt queues associated with link
398 * @l_ptr: pointer to link
400 void tipc_link_purge_queues(struct tipc_link *l_ptr)
402 kfree_skb_list(l_ptr->oldest_deferred_in);
403 kfree_skb_list(l_ptr->first_out);
404 tipc_link_reset_fragments(l_ptr);
405 kfree_skb(l_ptr->proto_msg_queue);
406 l_ptr->proto_msg_queue = NULL;
409 void tipc_link_reset(struct tipc_link *l_ptr)
411 u32 prev_state = l_ptr->state;
412 u32 checkpoint = l_ptr->next_in_no;
413 int was_active_link = tipc_link_is_active(l_ptr);
415 msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
417 /* Link is down, accept any session */
418 l_ptr->peer_session = INVALID_SESSION;
420 /* Prepare for max packet size negotiation */
421 link_init_max_pkt(l_ptr);
423 l_ptr->state = RESET_UNKNOWN;
425 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
428 tipc_node_link_down(l_ptr->owner, l_ptr);
429 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
431 if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
432 l_ptr->reset_checkpoint = checkpoint;
433 l_ptr->exp_msg_count = START_CHANGEOVER;
436 /* Clean up all queues: */
437 link_release_outqueue(l_ptr);
438 kfree_skb(l_ptr->proto_msg_queue);
439 l_ptr->proto_msg_queue = NULL;
440 kfree_skb_list(l_ptr->oldest_deferred_in);
441 if (!list_empty(&l_ptr->waiting_ports))
442 tipc_link_wakeup_ports(l_ptr, 1);
444 l_ptr->retransm_queue_head = 0;
445 l_ptr->retransm_queue_size = 0;
446 l_ptr->last_out = NULL;
447 l_ptr->first_out = NULL;
448 l_ptr->next_out = NULL;
449 l_ptr->unacked_window = 0;
450 l_ptr->checkpoint = 1;
451 l_ptr->next_out_no = 1;
452 l_ptr->deferred_inqueue_sz = 0;
453 l_ptr->oldest_deferred_in = NULL;
454 l_ptr->newest_deferred_in = NULL;
455 l_ptr->fsm_msg_cnt = 0;
456 l_ptr->stale_count = 0;
457 link_reset_statistics(l_ptr);
460 void tipc_link_reset_list(unsigned int bearer_id)
462 struct tipc_link *l_ptr;
463 struct tipc_node *n_ptr;
466 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
467 spin_lock_bh(&n_ptr->lock);
468 l_ptr = n_ptr->links[bearer_id];
470 tipc_link_reset(l_ptr);
471 spin_unlock_bh(&n_ptr->lock);
476 static void link_activate(struct tipc_link *l_ptr)
478 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
479 tipc_node_link_up(l_ptr->owner, l_ptr);
480 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
484 * link_state_event - link finite state machine
485 * @l_ptr: pointer to link
486 * @event: state machine event to process
488 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
490 struct tipc_link *other;
491 u32 cont_intv = l_ptr->continuity_interval;
493 if (l_ptr->flags & LINK_STOPPED)
496 if (!(l_ptr->flags & LINK_STARTED) && (event != STARTING_EVT))
497 return; /* Not yet. */
499 /* Check whether changeover is going on */
500 if (l_ptr->exp_msg_count) {
501 if (event == TIMEOUT_EVT)
502 link_set_timer(l_ptr, cont_intv);
506 switch (l_ptr->state) {
507 case WORKING_WORKING:
509 case TRAFFIC_MSG_EVT:
513 if (l_ptr->next_in_no != l_ptr->checkpoint) {
514 l_ptr->checkpoint = l_ptr->next_in_no;
515 if (tipc_bclink_acks_missing(l_ptr->owner)) {
516 tipc_link_proto_xmit(l_ptr, STATE_MSG,
518 l_ptr->fsm_msg_cnt++;
519 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
520 tipc_link_proto_xmit(l_ptr, STATE_MSG,
522 l_ptr->fsm_msg_cnt++;
524 link_set_timer(l_ptr, cont_intv);
527 l_ptr->state = WORKING_UNKNOWN;
528 l_ptr->fsm_msg_cnt = 0;
529 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
530 l_ptr->fsm_msg_cnt++;
531 link_set_timer(l_ptr, cont_intv / 4);
534 pr_info("%s<%s>, requested by peer\n", link_rst_msg,
536 tipc_link_reset(l_ptr);
537 l_ptr->state = RESET_RESET;
538 l_ptr->fsm_msg_cnt = 0;
539 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
541 l_ptr->fsm_msg_cnt++;
542 link_set_timer(l_ptr, cont_intv);
545 pr_err("%s%u in WW state\n", link_unk_evt, event);
548 case WORKING_UNKNOWN:
550 case TRAFFIC_MSG_EVT:
552 l_ptr->state = WORKING_WORKING;
553 l_ptr->fsm_msg_cnt = 0;
554 link_set_timer(l_ptr, cont_intv);
557 pr_info("%s<%s>, requested by peer while probing\n",
558 link_rst_msg, l_ptr->name);
559 tipc_link_reset(l_ptr);
560 l_ptr->state = RESET_RESET;
561 l_ptr->fsm_msg_cnt = 0;
562 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
564 l_ptr->fsm_msg_cnt++;
565 link_set_timer(l_ptr, cont_intv);
568 if (l_ptr->next_in_no != l_ptr->checkpoint) {
569 l_ptr->state = WORKING_WORKING;
570 l_ptr->fsm_msg_cnt = 0;
571 l_ptr->checkpoint = l_ptr->next_in_no;
572 if (tipc_bclink_acks_missing(l_ptr->owner)) {
573 tipc_link_proto_xmit(l_ptr, STATE_MSG,
575 l_ptr->fsm_msg_cnt++;
577 link_set_timer(l_ptr, cont_intv);
578 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
579 tipc_link_proto_xmit(l_ptr, STATE_MSG,
581 l_ptr->fsm_msg_cnt++;
582 link_set_timer(l_ptr, cont_intv / 4);
583 } else { /* Link has failed */
584 pr_warn("%s<%s>, peer not responding\n",
585 link_rst_msg, l_ptr->name);
586 tipc_link_reset(l_ptr);
587 l_ptr->state = RESET_UNKNOWN;
588 l_ptr->fsm_msg_cnt = 0;
589 tipc_link_proto_xmit(l_ptr, RESET_MSG,
591 l_ptr->fsm_msg_cnt++;
592 link_set_timer(l_ptr, cont_intv);
596 pr_err("%s%u in WU state\n", link_unk_evt, event);
601 case TRAFFIC_MSG_EVT:
604 other = l_ptr->owner->active_links[0];
605 if (other && link_working_unknown(other))
607 l_ptr->state = WORKING_WORKING;
608 l_ptr->fsm_msg_cnt = 0;
609 link_activate(l_ptr);
610 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
611 l_ptr->fsm_msg_cnt++;
612 if (l_ptr->owner->working_links == 1)
613 tipc_link_sync_xmit(l_ptr);
614 link_set_timer(l_ptr, cont_intv);
617 l_ptr->state = RESET_RESET;
618 l_ptr->fsm_msg_cnt = 0;
619 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
621 l_ptr->fsm_msg_cnt++;
622 link_set_timer(l_ptr, cont_intv);
625 l_ptr->flags |= LINK_STARTED;
628 tipc_link_proto_xmit(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
629 l_ptr->fsm_msg_cnt++;
630 link_set_timer(l_ptr, cont_intv);
633 pr_err("%s%u in RU state\n", link_unk_evt, event);
638 case TRAFFIC_MSG_EVT:
640 other = l_ptr->owner->active_links[0];
641 if (other && link_working_unknown(other))
643 l_ptr->state = WORKING_WORKING;
644 l_ptr->fsm_msg_cnt = 0;
645 link_activate(l_ptr);
646 tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
647 l_ptr->fsm_msg_cnt++;
648 if (l_ptr->owner->working_links == 1)
649 tipc_link_sync_xmit(l_ptr);
650 link_set_timer(l_ptr, cont_intv);
655 tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
657 l_ptr->fsm_msg_cnt++;
658 link_set_timer(l_ptr, cont_intv);
661 pr_err("%s%u in RR state\n", link_unk_evt, event);
665 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
670 * link_bundle_buf(): Append contents of a buffer to
671 * the tail of an existing one.
673 static int link_bundle_buf(struct tipc_link *l_ptr, struct sk_buff *bundler,
676 struct tipc_msg *bundler_msg = buf_msg(bundler);
677 struct tipc_msg *msg = buf_msg(buf);
678 u32 size = msg_size(msg);
679 u32 bundle_size = msg_size(bundler_msg);
680 u32 to_pos = align(bundle_size);
681 u32 pad = to_pos - bundle_size;
683 if (msg_user(bundler_msg) != MSG_BUNDLER)
685 if (msg_type(bundler_msg) != OPEN_MSG)
687 if (skb_tailroom(bundler) < (pad + size))
689 if (l_ptr->max_pkt < (to_pos + size))
692 skb_put(bundler, pad + size);
693 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
694 msg_set_size(bundler_msg, to_pos + size);
695 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
697 l_ptr->stats.sent_bundled++;
701 static void link_add_to_outqueue(struct tipc_link *l_ptr,
703 struct tipc_msg *msg)
705 u32 ack = mod(l_ptr->next_in_no - 1);
706 u32 seqno = mod(l_ptr->next_out_no++);
708 msg_set_word(msg, 2, ((ack << 16) | seqno));
709 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
711 if (l_ptr->first_out) {
712 l_ptr->last_out->next = buf;
713 l_ptr->last_out = buf;
715 l_ptr->first_out = l_ptr->last_out = buf;
717 l_ptr->out_queue_size++;
718 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
719 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
722 static void link_add_chain_to_outqueue(struct tipc_link *l_ptr,
723 struct sk_buff *buf_chain,
727 struct tipc_msg *msg;
729 if (!l_ptr->next_out)
730 l_ptr->next_out = buf_chain;
733 buf_chain = buf_chain->next;
736 msg_set_long_msgno(msg, long_msgno);
737 link_add_to_outqueue(l_ptr, buf, msg);
742 * tipc_link_xmit() is the 'full path' for messages, called from
743 * inside TIPC when the 'fast path' in tipc_send_xmit
744 * has failed, and from link_send()
746 int __tipc_link_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
748 struct tipc_msg *msg = buf_msg(buf);
749 u32 size = msg_size(msg);
750 u32 dsz = msg_data_sz(msg);
751 u32 queue_size = l_ptr->out_queue_size;
752 u32 imp = tipc_msg_tot_importance(msg);
753 u32 queue_limit = l_ptr->queue_limit[imp];
754 u32 max_packet = l_ptr->max_pkt;
756 /* Match msg importance against queue limits: */
757 if (unlikely(queue_size >= queue_limit)) {
758 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
759 link_schedule_port(l_ptr, msg_origport(msg), size);
764 if (imp > CONN_MANAGER) {
765 pr_warn("%s<%s>, send queue full", link_rst_msg,
767 tipc_link_reset(l_ptr);
772 /* Fragmentation needed ? */
773 if (size > max_packet)
774 return tipc_link_frag_xmit(l_ptr, buf);
776 /* Packet can be queued or sent. */
777 if (likely(!link_congested(l_ptr))) {
778 link_add_to_outqueue(l_ptr, buf, msg);
780 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
781 l_ptr->unacked_window = 0;
784 /* Congestion: can message be bundled ? */
785 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
786 (msg_user(msg) != MSG_FRAGMENTER)) {
788 /* Try adding message to an existing bundle */
789 if (l_ptr->next_out &&
790 link_bundle_buf(l_ptr, l_ptr->last_out, buf))
793 /* Try creating a new bundle */
794 if (size <= max_packet * 2 / 3) {
795 struct sk_buff *bundler = tipc_buf_acquire(max_packet);
796 struct tipc_msg bundler_hdr;
799 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
800 INT_H_SIZE, l_ptr->addr);
801 skb_copy_to_linear_data(bundler, &bundler_hdr,
803 skb_trim(bundler, INT_H_SIZE);
804 link_bundle_buf(l_ptr, bundler, buf);
807 l_ptr->stats.sent_bundles++;
811 if (!l_ptr->next_out)
812 l_ptr->next_out = buf;
813 link_add_to_outqueue(l_ptr, buf, msg);
818 * tipc_link_xmit(): same as __tipc_link_xmit(), but the link to use
819 * has not been selected yet, and the the owner node is not locked
820 * Called by TIPC internal users, e.g. the name distributor
822 int tipc_link_xmit(struct sk_buff *buf, u32 dest, u32 selector)
824 struct tipc_link *l_ptr;
825 struct tipc_node *n_ptr;
826 int res = -ELINKCONG;
828 read_lock_bh(&tipc_net_lock);
829 n_ptr = tipc_node_find(dest);
831 tipc_node_lock(n_ptr);
832 l_ptr = n_ptr->active_links[selector & 1];
834 res = __tipc_link_xmit(l_ptr, buf);
837 tipc_node_unlock(n_ptr);
841 read_unlock_bh(&tipc_net_lock);
846 * tipc_link_sync_xmit - synchronize broadcast link endpoints.
848 * Give a newly added peer node the sequence number where it should
849 * start receiving and acking broadcast packets.
851 * Called with node locked
853 static void tipc_link_sync_xmit(struct tipc_link *l)
856 struct tipc_msg *msg;
858 buf = tipc_buf_acquire(INT_H_SIZE);
863 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, l->addr);
864 msg_set_last_bcast(msg, l->owner->bclink.acked);
865 link_add_chain_to_outqueue(l, buf, 0);
866 tipc_link_push_queue(l);
870 * tipc_link_sync_rcv - synchronize broadcast link endpoints.
871 * Receive the sequence number where we should start receiving and
872 * acking broadcast packets from a newly added peer node, and open
873 * up for reception of such packets.
875 * Called with node locked
877 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf)
879 struct tipc_msg *msg = buf_msg(buf);
881 n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
882 n->bclink.recv_permitted = true;
887 * tipc_link_names_xmit - send name table entries to new neighbor
889 * Send routine for bulk delivery of name table messages when contact
890 * with a new neighbor occurs. No link congestion checking is performed
891 * because name table messages *must* be delivered. The messages must be
892 * small enough not to require fragmentation.
893 * Called without any locks held.
895 void tipc_link_names_xmit(struct list_head *message_list, u32 dest)
897 struct tipc_node *n_ptr;
898 struct tipc_link *l_ptr;
900 struct sk_buff *temp_buf;
902 if (list_empty(message_list))
905 read_lock_bh(&tipc_net_lock);
906 n_ptr = tipc_node_find(dest);
908 tipc_node_lock(n_ptr);
909 l_ptr = n_ptr->active_links[0];
911 /* convert circular list to linear list */
912 ((struct sk_buff *)message_list->prev)->next = NULL;
913 link_add_chain_to_outqueue(l_ptr,
914 (struct sk_buff *)message_list->next, 0);
915 tipc_link_push_queue(l_ptr);
916 INIT_LIST_HEAD(message_list);
918 tipc_node_unlock(n_ptr);
920 read_unlock_bh(&tipc_net_lock);
922 /* discard the messages if they couldn't be sent */
923 list_for_each_safe(buf, temp_buf, ((struct sk_buff *)message_list)) {
924 list_del((struct list_head *)buf);
930 * tipc_link_xmit_fast: Entry for data messages where the
931 * destination link is known and the header is complete,
932 * inclusive total message length. Very time critical.
933 * Link is locked. Returns user data length.
935 static int tipc_link_xmit_fast(struct tipc_link *l_ptr, struct sk_buff *buf,
938 struct tipc_msg *msg = buf_msg(buf);
939 int res = msg_data_sz(msg);
941 if (likely(!link_congested(l_ptr))) {
942 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
943 link_add_to_outqueue(l_ptr, buf, msg);
944 tipc_bearer_send(l_ptr->b_ptr, buf,
946 l_ptr->unacked_window = 0;
950 *used_max_pkt = l_ptr->max_pkt;
952 return __tipc_link_xmit(l_ptr, buf); /* All other cases */
956 * tipc_link_iovec_xmit_fast: Entry for messages where the
957 * destination processor is known and the header is complete,
958 * except for total message length.
959 * Returns user data length or errno.
961 int tipc_link_iovec_xmit_fast(struct tipc_port *sender,
962 struct iovec const *msg_sect,
963 unsigned int len, u32 destaddr)
965 struct tipc_msg *hdr = &sender->phdr;
966 struct tipc_link *l_ptr;
968 struct tipc_node *node;
970 u32 selector = msg_origport(hdr) & 1;
974 * Try building message using port's max_pkt hint.
975 * (Must not hold any locks while building message.)
977 res = tipc_msg_build(hdr, msg_sect, len, sender->max_pkt, &buf);
978 /* Exit if build request was invalid */
979 if (unlikely(res < 0))
982 read_lock_bh(&tipc_net_lock);
983 node = tipc_node_find(destaddr);
985 tipc_node_lock(node);
986 l_ptr = node->active_links[selector];
989 res = tipc_link_xmit_fast(l_ptr, buf,
992 tipc_node_unlock(node);
993 read_unlock_bh(&tipc_net_lock);
997 /* Exit if link (or bearer) is congested */
998 if (link_congested(l_ptr)) {
999 res = link_schedule_port(l_ptr,
1005 * Message size exceeds max_pkt hint; update hint,
1006 * then re-try fast path or fragment the message
1008 sender->max_pkt = l_ptr->max_pkt;
1009 tipc_node_unlock(node);
1010 read_unlock_bh(&tipc_net_lock);
1013 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1016 return tipc_link_iovec_long_xmit(sender, msg_sect,
1019 tipc_node_unlock(node);
1021 read_unlock_bh(&tipc_net_lock);
1023 /* Couldn't find a link to the destination node */
1025 tipc_port_iovec_reject(sender, hdr, msg_sect, len, TIPC_ERR_NO_NODE);
1026 return -ENETUNREACH;
1030 * tipc_link_iovec_long_xmit(): Entry for long messages where the
1031 * destination node is known and the header is complete,
1032 * inclusive total message length.
1033 * Link and bearer congestion status have been checked to be ok,
1034 * and are ignored if they change.
1036 * Note that fragments do not use the full link MTU so that they won't have
1037 * to undergo refragmentation if link changeover causes them to be sent
1038 * over another link with an additional tunnel header added as prefix.
1039 * (Refragmentation will still occur if the other link has a smaller MTU.)
1041 * Returns user data length or errno.
1043 static int tipc_link_iovec_long_xmit(struct tipc_port *sender,
1044 struct iovec const *msg_sect,
1045 unsigned int len, u32 destaddr)
1047 struct tipc_link *l_ptr;
1048 struct tipc_node *node;
1049 struct tipc_msg *hdr = &sender->phdr;
1051 u32 max_pkt, fragm_sz, rest;
1052 struct tipc_msg fragm_hdr;
1053 struct sk_buff *buf, *buf_chain, *prev;
1054 u32 fragm_crs, fragm_rest, hsz, sect_rest;
1055 const unchar __user *sect_crs;
1062 max_pkt = sender->max_pkt - INT_H_SIZE;
1063 /* leave room for tunnel header in case of link changeover */
1064 fragm_sz = max_pkt - INT_H_SIZE;
1065 /* leave room for fragmentation header in each fragment */
1073 /* Prepare reusable fragment header */
1074 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1075 INT_H_SIZE, msg_destnode(hdr));
1076 msg_set_size(&fragm_hdr, max_pkt);
1077 msg_set_fragm_no(&fragm_hdr, 1);
1079 /* Prepare header of first fragment */
1080 buf_chain = buf = tipc_buf_acquire(max_pkt);
1084 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1085 hsz = msg_hdr_sz(hdr);
1086 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1088 /* Chop up message */
1089 fragm_crs = INT_H_SIZE + hsz;
1090 fragm_rest = fragm_sz - hsz;
1092 do { /* For all sections */
1096 sect_rest = msg_sect[++curr_sect].iov_len;
1097 sect_crs = msg_sect[curr_sect].iov_base;
1100 if (sect_rest < fragm_rest)
1105 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1108 kfree_skb_list(buf_chain);
1117 if (!fragm_rest && rest) {
1119 /* Initiate new fragment: */
1120 if (rest <= fragm_sz) {
1122 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1124 msg_set_type(&fragm_hdr, FRAGMENT);
1126 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1127 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1129 buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1137 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1138 fragm_crs = INT_H_SIZE;
1139 fragm_rest = fragm_sz;
1144 * Now we have a buffer chain. Select a link and check
1145 * that packet size is still OK
1147 node = tipc_node_find(destaddr);
1149 tipc_node_lock(node);
1150 l_ptr = node->active_links[sender->ref & 1];
1152 tipc_node_unlock(node);
1155 if (l_ptr->max_pkt < max_pkt) {
1156 sender->max_pkt = l_ptr->max_pkt;
1157 tipc_node_unlock(node);
1158 kfree_skb_list(buf_chain);
1163 kfree_skb_list(buf_chain);
1164 tipc_port_iovec_reject(sender, hdr, msg_sect, len,
1166 return -ENETUNREACH;
1169 /* Append chain of fragments to send queue & send them */
1170 l_ptr->long_msg_seq_no++;
1171 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1172 l_ptr->stats.sent_fragments += fragm_no;
1173 l_ptr->stats.sent_fragmented++;
1174 tipc_link_push_queue(l_ptr);
1175 tipc_node_unlock(node);
1180 * tipc_link_push_packet: Push one unsent packet to the media
1182 static u32 tipc_link_push_packet(struct tipc_link *l_ptr)
1184 struct sk_buff *buf = l_ptr->first_out;
1185 u32 r_q_size = l_ptr->retransm_queue_size;
1186 u32 r_q_head = l_ptr->retransm_queue_head;
1188 /* Step to position where retransmission failed, if any, */
1189 /* consider that buffers may have been released in meantime */
1190 if (r_q_size && buf) {
1191 u32 last = lesser(mod(r_q_head + r_q_size),
1192 link_last_sent(l_ptr));
1193 u32 first = buf_seqno(buf);
1195 while (buf && less(first, r_q_head)) {
1196 first = mod(first + 1);
1199 l_ptr->retransm_queue_head = r_q_head = first;
1200 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1203 /* Continue retransmission now, if there is anything: */
1204 if (r_q_size && buf) {
1205 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1206 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1207 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1208 l_ptr->retransm_queue_head = mod(++r_q_head);
1209 l_ptr->retransm_queue_size = --r_q_size;
1210 l_ptr->stats.retransmitted++;
1214 /* Send deferred protocol message, if any: */
1215 buf = l_ptr->proto_msg_queue;
1217 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1218 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1219 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1220 l_ptr->unacked_window = 0;
1222 l_ptr->proto_msg_queue = NULL;
1226 /* Send one deferred data message, if send window not full: */
1227 buf = l_ptr->next_out;
1229 struct tipc_msg *msg = buf_msg(buf);
1230 u32 next = msg_seqno(msg);
1231 u32 first = buf_seqno(l_ptr->first_out);
1233 if (mod(next - first) < l_ptr->queue_limit[0]) {
1234 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1235 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1236 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1237 if (msg_user(msg) == MSG_BUNDLER)
1238 msg_set_type(msg, CLOSED_MSG);
1239 l_ptr->next_out = buf->next;
1247 * push_queue(): push out the unsent messages of a link where
1248 * congestion has abated. Node is locked
1250 void tipc_link_push_queue(struct tipc_link *l_ptr)
1255 res = tipc_link_push_packet(l_ptr);
1259 static void link_reset_all(unsigned long addr)
1261 struct tipc_node *n_ptr;
1262 char addr_string[16];
1265 read_lock_bh(&tipc_net_lock);
1266 n_ptr = tipc_node_find((u32)addr);
1268 read_unlock_bh(&tipc_net_lock);
1269 return; /* node no longer exists */
1272 tipc_node_lock(n_ptr);
1274 pr_warn("Resetting all links to %s\n",
1275 tipc_addr_string_fill(addr_string, n_ptr->addr));
1277 for (i = 0; i < MAX_BEARERS; i++) {
1278 if (n_ptr->links[i]) {
1279 link_print(n_ptr->links[i], "Resetting link\n");
1280 tipc_link_reset(n_ptr->links[i]);
1284 tipc_node_unlock(n_ptr);
1285 read_unlock_bh(&tipc_net_lock);
1288 static void link_retransmit_failure(struct tipc_link *l_ptr,
1289 struct sk_buff *buf)
1291 struct tipc_msg *msg = buf_msg(buf);
1293 pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
1296 /* Handle failure on standard link */
1297 link_print(l_ptr, "Resetting link\n");
1298 tipc_link_reset(l_ptr);
1301 /* Handle failure on broadcast link */
1302 struct tipc_node *n_ptr;
1303 char addr_string[16];
1305 pr_info("Msg seq number: %u, ", msg_seqno(msg));
1306 pr_cont("Outstanding acks: %lu\n",
1307 (unsigned long) TIPC_SKB_CB(buf)->handle);
1309 n_ptr = tipc_bclink_retransmit_to();
1310 tipc_node_lock(n_ptr);
1312 tipc_addr_string_fill(addr_string, n_ptr->addr);
1313 pr_info("Broadcast link info for %s\n", addr_string);
1314 pr_info("Reception permitted: %d, Acked: %u\n",
1315 n_ptr->bclink.recv_permitted,
1316 n_ptr->bclink.acked);
1317 pr_info("Last in: %u, Oos state: %u, Last sent: %u\n",
1318 n_ptr->bclink.last_in,
1319 n_ptr->bclink.oos_state,
1320 n_ptr->bclink.last_sent);
1322 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1324 tipc_node_unlock(n_ptr);
1326 l_ptr->stale_count = 0;
1330 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
1333 struct tipc_msg *msg;
1340 /* Detect repeated retransmit failures */
1341 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1342 if (++l_ptr->stale_count > 100) {
1343 link_retransmit_failure(l_ptr, buf);
1347 l_ptr->last_retransmitted = msg_seqno(msg);
1348 l_ptr->stale_count = 1;
1351 while (retransmits && (buf != l_ptr->next_out) && buf) {
1353 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1354 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1355 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1358 l_ptr->stats.retransmitted++;
1361 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1365 * link_insert_deferred_queue - insert deferred messages back into receive chain
1367 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1368 struct sk_buff *buf)
1372 if (l_ptr->oldest_deferred_in == NULL)
1375 seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1376 if (seq_no == mod(l_ptr->next_in_no)) {
1377 l_ptr->newest_deferred_in->next = buf;
1378 buf = l_ptr->oldest_deferred_in;
1379 l_ptr->oldest_deferred_in = NULL;
1380 l_ptr->deferred_inqueue_sz = 0;
1386 * link_recv_buf_validate - validate basic format of received message
1388 * This routine ensures a TIPC message has an acceptable header, and at least
1389 * as much data as the header indicates it should. The routine also ensures
1390 * that the entire message header is stored in the main fragment of the message
1391 * buffer, to simplify future access to message header fields.
1393 * Note: Having extra info present in the message header or data areas is OK.
1394 * TIPC will ignore the excess, under the assumption that it is optional info
1395 * introduced by a later release of the protocol.
1397 static int link_recv_buf_validate(struct sk_buff *buf)
1399 static u32 min_data_hdr_size[8] = {
1400 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1401 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1404 struct tipc_msg *msg;
1410 /* If this packet comes from the defer queue, the skb has already
1413 if (unlikely(TIPC_SKB_CB(buf)->deferred))
1416 if (unlikely(buf->len < MIN_H_SIZE))
1419 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1423 if (unlikely(msg_version(msg) != TIPC_VERSION))
1426 size = msg_size(msg);
1427 hdr_size = msg_hdr_sz(msg);
1428 min_hdr_size = msg_isdata(msg) ?
1429 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1431 if (unlikely((hdr_size < min_hdr_size) ||
1432 (size < hdr_size) ||
1433 (buf->len < size) ||
1434 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1437 return pskb_may_pull(buf, hdr_size);
1441 * tipc_rcv - process TIPC packets/messages arriving from off-node
1442 * @head: pointer to message buffer chain
1443 * @tb_ptr: pointer to bearer message arrived on
1445 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1446 * structure (i.e. cannot be NULL), but bearer can be inactive.
1448 void tipc_rcv(struct sk_buff *head, struct tipc_bearer *b_ptr)
1450 read_lock_bh(&tipc_net_lock);
1452 struct tipc_node *n_ptr;
1453 struct tipc_link *l_ptr;
1454 struct sk_buff *crs;
1455 struct sk_buff *buf = head;
1456 struct tipc_msg *msg;
1464 /* Ensure message is well-formed */
1465 if (unlikely(!link_recv_buf_validate(buf)))
1468 /* Ensure message data is a single contiguous unit */
1469 if (unlikely(skb_linearize(buf)))
1472 /* Handle arrival of a non-unicast link message */
1475 if (unlikely(msg_non_seq(msg))) {
1476 if (msg_user(msg) == LINK_CONFIG)
1477 tipc_disc_rcv(buf, b_ptr);
1479 tipc_bclink_rcv(buf);
1483 /* Discard unicast link messages destined for another node */
1484 if (unlikely(!msg_short(msg) &&
1485 (msg_destnode(msg) != tipc_own_addr)))
1488 /* Locate neighboring node that sent message */
1489 n_ptr = tipc_node_find(msg_prevnode(msg));
1490 if (unlikely(!n_ptr))
1492 tipc_node_lock(n_ptr);
1494 /* Locate unicast link endpoint that should handle message */
1495 l_ptr = n_ptr->links[b_ptr->identity];
1496 if (unlikely(!l_ptr))
1497 goto unlock_discard;
1499 /* Verify that communication with node is currently allowed */
1500 if ((n_ptr->block_setup & WAIT_PEER_DOWN) &&
1501 msg_user(msg) == LINK_PROTOCOL &&
1502 (msg_type(msg) == RESET_MSG ||
1503 msg_type(msg) == ACTIVATE_MSG) &&
1504 !msg_redundant_link(msg))
1505 n_ptr->block_setup &= ~WAIT_PEER_DOWN;
1507 if (n_ptr->block_setup)
1508 goto unlock_discard;
1510 /* Validate message sequence number info */
1511 seq_no = msg_seqno(msg);
1512 ackd = msg_ack(msg);
1514 /* Release acked messages */
1515 if (n_ptr->bclink.recv_permitted)
1516 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1518 crs = l_ptr->first_out;
1519 while ((crs != l_ptr->next_out) &&
1520 less_eq(buf_seqno(crs), ackd)) {
1521 struct sk_buff *next = crs->next;
1527 l_ptr->first_out = crs;
1528 l_ptr->out_queue_size -= released;
1531 /* Try sending any messages link endpoint has pending */
1532 if (unlikely(l_ptr->next_out))
1533 tipc_link_push_queue(l_ptr);
1535 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1536 tipc_link_wakeup_ports(l_ptr, 0);
1538 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1539 l_ptr->stats.sent_acks++;
1540 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1543 /* Process the incoming packet */
1544 if (unlikely(!link_working_working(l_ptr))) {
1545 if (msg_user(msg) == LINK_PROTOCOL) {
1546 tipc_link_proto_rcv(l_ptr, buf);
1547 head = link_insert_deferred_queue(l_ptr, head);
1548 tipc_node_unlock(n_ptr);
1552 /* Traffic message. Conditionally activate link */
1553 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1555 if (link_working_working(l_ptr)) {
1556 /* Re-insert buffer in front of queue */
1559 tipc_node_unlock(n_ptr);
1562 goto unlock_discard;
1565 /* Link is now in state WORKING_WORKING */
1566 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1567 link_handle_out_of_seq_msg(l_ptr, buf);
1568 head = link_insert_deferred_queue(l_ptr, head);
1569 tipc_node_unlock(n_ptr);
1572 l_ptr->next_in_no++;
1573 if (unlikely(l_ptr->oldest_deferred_in))
1574 head = link_insert_deferred_queue(l_ptr, head);
1576 /* Deliver packet/message to correct user: */
1577 if (unlikely(msg_user(msg) == CHANGEOVER_PROTOCOL)) {
1578 if (!tipc_link_tunnel_rcv(n_ptr, &buf)) {
1579 tipc_node_unlock(n_ptr);
1583 } else if (msg_user(msg) == MSG_FRAGMENTER) {
1586 l_ptr->stats.recv_fragments++;
1587 rc = tipc_link_frag_rcv(&l_ptr->reasm_head,
1590 if (rc == LINK_REASM_COMPLETE) {
1591 l_ptr->stats.recv_fragmented++;
1594 if (rc == LINK_REASM_ERROR)
1595 tipc_link_reset(l_ptr);
1596 tipc_node_unlock(n_ptr);
1601 switch (msg_user(msg)) {
1602 case TIPC_LOW_IMPORTANCE:
1603 case TIPC_MEDIUM_IMPORTANCE:
1604 case TIPC_HIGH_IMPORTANCE:
1605 case TIPC_CRITICAL_IMPORTANCE:
1606 tipc_node_unlock(n_ptr);
1610 l_ptr->stats.recv_bundles++;
1611 l_ptr->stats.recv_bundled += msg_msgcnt(msg);
1612 tipc_node_unlock(n_ptr);
1613 tipc_link_bundle_rcv(buf);
1615 case NAME_DISTRIBUTOR:
1616 n_ptr->bclink.recv_permitted = true;
1617 tipc_node_unlock(n_ptr);
1618 tipc_named_rcv(buf);
1621 tipc_node_unlock(n_ptr);
1622 tipc_port_proto_rcv(buf);
1624 case BCAST_PROTOCOL:
1625 tipc_link_sync_rcv(n_ptr, buf);
1631 tipc_node_unlock(n_ptr);
1634 tipc_node_unlock(n_ptr);
1638 read_unlock_bh(&tipc_net_lock);
1642 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1644 * Returns increase in queue length (i.e. 0 or 1)
1646 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1647 struct sk_buff *buf)
1649 struct sk_buff *queue_buf;
1650 struct sk_buff **prev;
1651 u32 seq_no = buf_seqno(buf);
1656 if (*head == NULL) {
1657 *head = *tail = buf;
1662 if (less(buf_seqno(*tail), seq_no)) {
1663 (*tail)->next = buf;
1668 /* Locate insertion point in queue, then insert; discard if duplicate */
1672 u32 curr_seqno = buf_seqno(queue_buf);
1674 if (seq_no == curr_seqno) {
1679 if (less(seq_no, curr_seqno))
1682 prev = &queue_buf->next;
1683 queue_buf = queue_buf->next;
1686 buf->next = queue_buf;
1692 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1694 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1695 struct sk_buff *buf)
1697 u32 seq_no = buf_seqno(buf);
1699 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1700 tipc_link_proto_rcv(l_ptr, buf);
1704 /* Record OOS packet arrival (force mismatch on next timeout) */
1705 l_ptr->checkpoint--;
1708 * Discard packet if a duplicate; otherwise add it to deferred queue
1709 * and notify peer of gap as per protocol specification
1711 if (less(seq_no, mod(l_ptr->next_in_no))) {
1712 l_ptr->stats.duplicates++;
1717 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1718 &l_ptr->newest_deferred_in, buf)) {
1719 l_ptr->deferred_inqueue_sz++;
1720 l_ptr->stats.deferred_recv++;
1721 TIPC_SKB_CB(buf)->deferred = true;
1722 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1723 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1725 l_ptr->stats.duplicates++;
1729 * Send protocol message to the other endpoint.
1731 void tipc_link_proto_xmit(struct tipc_link *l_ptr, u32 msg_typ, int probe_msg,
1732 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1734 struct sk_buff *buf = NULL;
1735 struct tipc_msg *msg = l_ptr->pmsg;
1736 u32 msg_size = sizeof(l_ptr->proto_msg);
1739 /* Discard any previous message that was deferred due to congestion */
1740 if (l_ptr->proto_msg_queue) {
1741 kfree_skb(l_ptr->proto_msg_queue);
1742 l_ptr->proto_msg_queue = NULL;
1745 /* Don't send protocol message during link changeover */
1746 if (l_ptr->exp_msg_count)
1749 /* Abort non-RESET send if communication with node is prohibited */
1750 if ((l_ptr->owner->block_setup) && (msg_typ != RESET_MSG))
1753 /* Create protocol message with "out-of-sequence" sequence number */
1754 msg_set_type(msg, msg_typ);
1755 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1756 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1757 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1759 if (msg_typ == STATE_MSG) {
1760 u32 next_sent = mod(l_ptr->next_out_no);
1762 if (!tipc_link_is_up(l_ptr))
1764 if (l_ptr->next_out)
1765 next_sent = buf_seqno(l_ptr->next_out);
1766 msg_set_next_sent(msg, next_sent);
1767 if (l_ptr->oldest_deferred_in) {
1768 u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1769 gap = mod(rec - mod(l_ptr->next_in_no));
1771 msg_set_seq_gap(msg, gap);
1773 l_ptr->stats.sent_nacks++;
1774 msg_set_link_tolerance(msg, tolerance);
1775 msg_set_linkprio(msg, priority);
1776 msg_set_max_pkt(msg, ack_mtu);
1777 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1778 msg_set_probe(msg, probe_msg != 0);
1780 u32 mtu = l_ptr->max_pkt;
1782 if ((mtu < l_ptr->max_pkt_target) &&
1783 link_working_working(l_ptr) &&
1784 l_ptr->fsm_msg_cnt) {
1785 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1786 if (l_ptr->max_pkt_probes == 10) {
1787 l_ptr->max_pkt_target = (msg_size - 4);
1788 l_ptr->max_pkt_probes = 0;
1789 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1791 l_ptr->max_pkt_probes++;
1794 l_ptr->stats.sent_probes++;
1796 l_ptr->stats.sent_states++;
1797 } else { /* RESET_MSG or ACTIVATE_MSG */
1798 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1799 msg_set_seq_gap(msg, 0);
1800 msg_set_next_sent(msg, 1);
1801 msg_set_probe(msg, 0);
1802 msg_set_link_tolerance(msg, l_ptr->tolerance);
1803 msg_set_linkprio(msg, l_ptr->priority);
1804 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1807 r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1808 msg_set_redundant_link(msg, r_flag);
1809 msg_set_linkprio(msg, l_ptr->priority);
1810 msg_set_size(msg, msg_size);
1812 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1814 buf = tipc_buf_acquire(msg_size);
1818 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1819 buf->priority = TC_PRIO_CONTROL;
1821 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1822 l_ptr->unacked_window = 0;
1827 * Receive protocol message :
1828 * Note that network plane id propagates through the network, and may
1829 * change at any time. The node with lowest address rules
1831 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf)
1837 struct tipc_msg *msg = buf_msg(buf);
1839 /* Discard protocol message during link changeover */
1840 if (l_ptr->exp_msg_count)
1843 /* record unnumbered packet arrival (force mismatch on next timeout) */
1844 l_ptr->checkpoint--;
1846 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
1847 if (tipc_own_addr > msg_prevnode(msg))
1848 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
1850 switch (msg_type(msg)) {
1853 if (!link_working_unknown(l_ptr) &&
1854 (l_ptr->peer_session != INVALID_SESSION)) {
1855 if (less_eq(msg_session(msg), l_ptr->peer_session))
1856 break; /* duplicate or old reset: ignore */
1859 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1860 link_working_unknown(l_ptr))) {
1862 * peer has lost contact -- don't allow peer's links
1863 * to reactivate before we recognize loss & clean up
1865 l_ptr->owner->block_setup = WAIT_NODE_DOWN;
1868 link_state_event(l_ptr, RESET_MSG);
1872 /* Update link settings according other endpoint's values */
1873 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1875 msg_tol = msg_link_tolerance(msg);
1876 if (msg_tol > l_ptr->tolerance)
1877 link_set_supervision_props(l_ptr, msg_tol);
1879 if (msg_linkprio(msg) > l_ptr->priority)
1880 l_ptr->priority = msg_linkprio(msg);
1882 max_pkt_info = msg_max_pkt(msg);
1884 if (max_pkt_info < l_ptr->max_pkt_target)
1885 l_ptr->max_pkt_target = max_pkt_info;
1886 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1887 l_ptr->max_pkt = l_ptr->max_pkt_target;
1889 l_ptr->max_pkt = l_ptr->max_pkt_target;
1892 /* Synchronize broadcast link info, if not done previously */
1893 if (!tipc_node_is_up(l_ptr->owner)) {
1894 l_ptr->owner->bclink.last_sent =
1895 l_ptr->owner->bclink.last_in =
1896 msg_last_bcast(msg);
1897 l_ptr->owner->bclink.oos_state = 0;
1900 l_ptr->peer_session = msg_session(msg);
1901 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1903 if (msg_type(msg) == ACTIVATE_MSG)
1904 link_state_event(l_ptr, ACTIVATE_MSG);
1908 msg_tol = msg_link_tolerance(msg);
1910 link_set_supervision_props(l_ptr, msg_tol);
1912 if (msg_linkprio(msg) &&
1913 (msg_linkprio(msg) != l_ptr->priority)) {
1914 pr_warn("%s<%s>, priority change %u->%u\n",
1915 link_rst_msg, l_ptr->name, l_ptr->priority,
1917 l_ptr->priority = msg_linkprio(msg);
1918 tipc_link_reset(l_ptr); /* Enforce change to take effect */
1921 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1922 l_ptr->stats.recv_states++;
1923 if (link_reset_unknown(l_ptr))
1926 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1927 rec_gap = mod(msg_next_sent(msg) -
1928 mod(l_ptr->next_in_no));
1931 max_pkt_ack = msg_max_pkt(msg);
1932 if (max_pkt_ack > l_ptr->max_pkt) {
1933 l_ptr->max_pkt = max_pkt_ack;
1934 l_ptr->max_pkt_probes = 0;
1938 if (msg_probe(msg)) {
1939 l_ptr->stats.recv_probes++;
1940 if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1941 max_pkt_ack = msg_size(msg);
1944 /* Protocol message before retransmits, reduce loss risk */
1945 if (l_ptr->owner->bclink.recv_permitted)
1946 tipc_bclink_update_link_state(l_ptr->owner,
1947 msg_last_bcast(msg));
1949 if (rec_gap || (msg_probe(msg))) {
1950 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, rec_gap, 0,
1953 if (msg_seq_gap(msg)) {
1954 l_ptr->stats.recv_nacks++;
1955 tipc_link_retransmit(l_ptr, l_ptr->first_out,
1965 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1966 * a different bearer. Owner node is locked.
1968 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1969 struct tipc_msg *tunnel_hdr,
1970 struct tipc_msg *msg,
1973 struct tipc_link *tunnel;
1974 struct sk_buff *buf;
1975 u32 length = msg_size(msg);
1977 tunnel = l_ptr->owner->active_links[selector & 1];
1978 if (!tipc_link_is_up(tunnel)) {
1979 pr_warn("%stunnel link no longer available\n", link_co_err);
1982 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1983 buf = tipc_buf_acquire(length + INT_H_SIZE);
1985 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1988 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
1989 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
1990 __tipc_link_xmit(tunnel, buf);
1994 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1995 * link is still active. We can do failover. Tunnel the failing link's
1996 * whole send queue via the remaining link. This way, we don't lose
1997 * any packets, and sequence order is preserved for subsequent traffic
1998 * sent over the remaining link. Owner node is locked.
2000 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
2002 u32 msgcount = l_ptr->out_queue_size;
2003 struct sk_buff *crs = l_ptr->first_out;
2004 struct tipc_link *tunnel = l_ptr->owner->active_links[0];
2005 struct tipc_msg tunnel_hdr;
2011 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2012 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2013 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2014 msg_set_msgcnt(&tunnel_hdr, msgcount);
2016 if (!l_ptr->first_out) {
2017 struct sk_buff *buf;
2019 buf = tipc_buf_acquire(INT_H_SIZE);
2021 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2022 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2023 __tipc_link_xmit(tunnel, buf);
2025 pr_warn("%sunable to send changeover msg\n",
2031 split_bundles = (l_ptr->owner->active_links[0] !=
2032 l_ptr->owner->active_links[1]);
2035 struct tipc_msg *msg = buf_msg(crs);
2037 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2038 struct tipc_msg *m = msg_get_wrapped(msg);
2039 unchar *pos = (unchar *)m;
2041 msgcount = msg_msgcnt(msg);
2042 while (msgcount--) {
2043 msg_set_seqno(m, msg_seqno(msg));
2044 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
2045 msg_link_selector(m));
2046 pos += align(msg_size(m));
2047 m = (struct tipc_msg *)pos;
2050 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
2051 msg_link_selector(msg));
2057 /* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
2058 * duplicate of the first link's send queue via the new link. This way, we
2059 * are guaranteed that currently queued packets from a socket are delivered
2060 * before future traffic from the same socket, even if this is using the
2061 * new link. The last arriving copy of each duplicate packet is dropped at
2062 * the receiving end by the regular protocol check, so packet cardinality
2063 * and sequence order is preserved per sender/receiver socket pair.
2064 * Owner node is locked.
2066 void tipc_link_dup_queue_xmit(struct tipc_link *l_ptr,
2067 struct tipc_link *tunnel)
2069 struct sk_buff *iter;
2070 struct tipc_msg tunnel_hdr;
2072 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2073 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2074 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2075 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2076 iter = l_ptr->first_out;
2078 struct sk_buff *outbuf;
2079 struct tipc_msg *msg = buf_msg(iter);
2080 u32 length = msg_size(msg);
2082 if (msg_user(msg) == MSG_BUNDLER)
2083 msg_set_type(msg, CLOSED_MSG);
2084 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2085 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2086 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2087 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2088 if (outbuf == NULL) {
2089 pr_warn("%sunable to send duplicate msg\n",
2093 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2094 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2096 __tipc_link_xmit(tunnel, outbuf);
2097 if (!tipc_link_is_up(l_ptr))
2104 * buf_extract - extracts embedded TIPC message from another message
2105 * @skb: encapsulating message buffer
2106 * @from_pos: offset to extract from
2108 * Returns a new message buffer containing an embedded message. The
2109 * encapsulating message itself is left unchanged.
2111 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2113 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2114 u32 size = msg_size(msg);
2117 eb = tipc_buf_acquire(size);
2119 skb_copy_to_linear_data(eb, msg, size);
2125 /* tipc_link_dup_rcv(): Receive a tunnelled DUPLICATE_MSG packet.
2126 * Owner node is locked.
2128 static void tipc_link_dup_rcv(struct tipc_link *l_ptr,
2129 struct sk_buff *t_buf)
2131 struct sk_buff *buf;
2133 if (!tipc_link_is_up(l_ptr))
2136 buf = buf_extract(t_buf, INT_H_SIZE);
2138 pr_warn("%sfailed to extract inner dup pkt\n", link_co_err);
2142 /* Add buffer to deferred queue, if applicable: */
2143 link_handle_out_of_seq_msg(l_ptr, buf);
2146 /* tipc_link_failover_rcv(): Receive a tunnelled ORIGINAL_MSG packet
2147 * Owner node is locked.
2149 static struct sk_buff *tipc_link_failover_rcv(struct tipc_link *l_ptr,
2150 struct sk_buff *t_buf)
2152 struct tipc_msg *t_msg = buf_msg(t_buf);
2153 struct sk_buff *buf = NULL;
2154 struct tipc_msg *msg;
2156 if (tipc_link_is_up(l_ptr))
2157 tipc_link_reset(l_ptr);
2159 /* First failover packet? */
2160 if (l_ptr->exp_msg_count == START_CHANGEOVER)
2161 l_ptr->exp_msg_count = msg_msgcnt(t_msg);
2163 /* Should there be an inner packet? */
2164 if (l_ptr->exp_msg_count) {
2165 l_ptr->exp_msg_count--;
2166 buf = buf_extract(t_buf, INT_H_SIZE);
2168 pr_warn("%sno inner failover pkt\n", link_co_err);
2173 if (less(msg_seqno(msg), l_ptr->reset_checkpoint)) {
2178 if (msg_user(msg) == MSG_FRAGMENTER) {
2179 l_ptr->stats.recv_fragments++;
2180 tipc_link_frag_rcv(&l_ptr->reasm_head,
2186 if ((l_ptr->exp_msg_count == 0) && (l_ptr->flags & LINK_STOPPED)) {
2187 tipc_node_detach_link(l_ptr->owner, l_ptr);
2193 /* tipc_link_tunnel_rcv(): Receive a tunnelled packet, sent
2194 * via other link as result of a failover (ORIGINAL_MSG) or
2195 * a new active link (DUPLICATE_MSG). Failover packets are
2196 * returned to the active link for delivery upwards.
2197 * Owner node is locked.
2199 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
2200 struct sk_buff **buf)
2202 struct sk_buff *t_buf = *buf;
2203 struct tipc_link *l_ptr;
2204 struct tipc_msg *t_msg = buf_msg(t_buf);
2205 u32 bearer_id = msg_bearer_id(t_msg);
2209 if (bearer_id >= MAX_BEARERS)
2212 l_ptr = n_ptr->links[bearer_id];
2216 if (msg_type(t_msg) == DUPLICATE_MSG)
2217 tipc_link_dup_rcv(l_ptr, t_buf);
2218 else if (msg_type(t_msg) == ORIGINAL_MSG)
2219 *buf = tipc_link_failover_rcv(l_ptr, t_buf);
2221 pr_warn("%sunknown tunnel pkt received\n", link_co_err);
2224 return *buf != NULL;
2228 * Bundler functionality:
2230 void tipc_link_bundle_rcv(struct sk_buff *buf)
2232 u32 msgcount = msg_msgcnt(buf_msg(buf));
2233 u32 pos = INT_H_SIZE;
2234 struct sk_buff *obuf;
2236 while (msgcount--) {
2237 obuf = buf_extract(buf, pos);
2239 pr_warn("Link unable to unbundle message(s)\n");
2242 pos += align(msg_size(buf_msg(obuf)));
2243 tipc_net_route_msg(obuf);
2249 * Fragmentation/defragmentation:
2253 * tipc_link_frag_xmit: Entry for buffers needing fragmentation.
2254 * The buffer is complete, inclusive total message length.
2255 * Returns user data length.
2257 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
2259 struct sk_buff *buf_chain = NULL;
2260 struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2261 struct tipc_msg *inmsg = buf_msg(buf);
2262 struct tipc_msg fragm_hdr;
2263 u32 insize = msg_size(inmsg);
2264 u32 dsz = msg_data_sz(inmsg);
2265 unchar *crs = buf->data;
2267 u32 pack_sz = l_ptr->max_pkt;
2268 u32 fragm_sz = pack_sz - INT_H_SIZE;
2272 if (msg_short(inmsg))
2273 destaddr = l_ptr->addr;
2275 destaddr = msg_destnode(inmsg);
2277 /* Prepare reusable fragment header: */
2278 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2279 INT_H_SIZE, destaddr);
2281 /* Chop up message: */
2283 struct sk_buff *fragm;
2285 if (rest <= fragm_sz) {
2287 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2289 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2290 if (fragm == NULL) {
2292 kfree_skb_list(buf_chain);
2295 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2297 msg_set_fragm_no(&fragm_hdr, fragm_no);
2298 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2299 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2301 buf_chain_tail->next = fragm;
2302 buf_chain_tail = fragm;
2306 msg_set_type(&fragm_hdr, FRAGMENT);
2310 /* Append chain of fragments to send queue & send them */
2311 l_ptr->long_msg_seq_no++;
2312 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2313 l_ptr->stats.sent_fragments += fragm_no;
2314 l_ptr->stats.sent_fragmented++;
2315 tipc_link_push_queue(l_ptr);
2320 /* tipc_link_frag_rcv(): Called with node lock on. Returns
2321 * the reassembled buffer if message is complete.
2323 int tipc_link_frag_rcv(struct sk_buff **head, struct sk_buff **tail,
2324 struct sk_buff **fbuf)
2326 struct sk_buff *frag = *fbuf;
2327 struct tipc_msg *msg = buf_msg(frag);
2328 u32 fragid = msg_type(msg);
2332 skb_pull(frag, msg_hdr_sz(msg));
2333 if (fragid == FIRST_FRAGMENT) {
2334 if (*head || skb_unclone(frag, GFP_ATOMIC))
2337 skb_frag_list_init(*head);
2341 skb_try_coalesce(*head, frag, &headstolen, &delta)) {
2342 kfree_skb_partial(frag, headstolen);
2346 if (!skb_has_frag_list(*head))
2347 skb_shinfo(*head)->frag_list = frag;
2349 (*tail)->next = frag;
2351 (*head)->truesize += frag->truesize;
2353 if (fragid == LAST_FRAGMENT) {
2355 *tail = *head = NULL;
2356 return LINK_REASM_COMPLETE;
2361 pr_warn_ratelimited("Link unable to reassemble fragmented message\n");
2364 return LINK_REASM_ERROR;
2367 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
2369 if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2372 l_ptr->tolerance = tolerance;
2373 l_ptr->continuity_interval =
2374 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2375 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2378 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
2380 /* Data messages from this node, inclusive FIRST_FRAGM */
2381 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2382 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2383 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2384 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2385 /* Transiting data messages,inclusive FIRST_FRAGM */
2386 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2387 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2388 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2389 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2390 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2391 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2392 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2393 /* FRAGMENT and LAST_FRAGMENT packets */
2394 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2397 /* tipc_link_find_owner - locate owner node of link by link's name
2398 * @name: pointer to link name string
2399 * @bearer_id: pointer to index in 'node->links' array where the link was found.
2400 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2401 * this also prevents link deletion.
2403 * Returns pointer to node owning the link, or 0 if no matching link is found.
2405 static struct tipc_node *tipc_link_find_owner(const char *link_name,
2406 unsigned int *bearer_id)
2408 struct tipc_link *l_ptr;
2409 struct tipc_node *n_ptr;
2410 struct tipc_node *found_node = 0;
2415 list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
2416 tipc_node_lock(n_ptr);
2417 for (i = 0; i < MAX_BEARERS; i++) {
2418 l_ptr = n_ptr->links[i];
2419 if (l_ptr && !strcmp(l_ptr->name, link_name)) {
2425 tipc_node_unlock(n_ptr);
2435 * link_value_is_valid -- validate proposed link tolerance/priority/window
2437 * @cmd: value type (TIPC_CMD_SET_LINK_*)
2438 * @new_value: the new value
2440 * Returns 1 if value is within range, 0 if not.
2442 static int link_value_is_valid(u16 cmd, u32 new_value)
2445 case TIPC_CMD_SET_LINK_TOL:
2446 return (new_value >= TIPC_MIN_LINK_TOL) &&
2447 (new_value <= TIPC_MAX_LINK_TOL);
2448 case TIPC_CMD_SET_LINK_PRI:
2449 return (new_value <= TIPC_MAX_LINK_PRI);
2450 case TIPC_CMD_SET_LINK_WINDOW:
2451 return (new_value >= TIPC_MIN_LINK_WIN) &&
2452 (new_value <= TIPC_MAX_LINK_WIN);
2458 * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2459 * @name: ptr to link, bearer, or media name
2460 * @new_value: new value of link, bearer, or media setting
2461 * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2463 * Caller must hold 'tipc_net_lock' to ensure link/bearer/media is not deleted.
2465 * Returns 0 if value updated and negative value on error.
2467 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2469 struct tipc_node *node;
2470 struct tipc_link *l_ptr;
2471 struct tipc_bearer *b_ptr;
2472 struct tipc_media *m_ptr;
2476 node = tipc_link_find_owner(name, &bearer_id);
2478 tipc_node_lock(node);
2479 l_ptr = node->links[bearer_id];
2483 case TIPC_CMD_SET_LINK_TOL:
2484 link_set_supervision_props(l_ptr, new_value);
2485 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2488 case TIPC_CMD_SET_LINK_PRI:
2489 l_ptr->priority = new_value;
2490 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2493 case TIPC_CMD_SET_LINK_WINDOW:
2494 tipc_link_set_queue_limits(l_ptr, new_value);
2501 tipc_node_unlock(node);
2505 b_ptr = tipc_bearer_find(name);
2508 case TIPC_CMD_SET_LINK_TOL:
2509 b_ptr->tolerance = new_value;
2511 case TIPC_CMD_SET_LINK_PRI:
2512 b_ptr->priority = new_value;
2514 case TIPC_CMD_SET_LINK_WINDOW:
2515 b_ptr->window = new_value;
2524 m_ptr = tipc_media_find(name);
2528 case TIPC_CMD_SET_LINK_TOL:
2529 m_ptr->tolerance = new_value;
2531 case TIPC_CMD_SET_LINK_PRI:
2532 m_ptr->priority = new_value;
2534 case TIPC_CMD_SET_LINK_WINDOW:
2535 m_ptr->window = new_value;
2544 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2547 struct tipc_link_config *args;
2551 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2552 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2554 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2555 new_value = ntohl(args->value);
2557 if (!link_value_is_valid(cmd, new_value))
2558 return tipc_cfg_reply_error_string(
2559 "cannot change, value invalid");
2561 if (!strcmp(args->name, tipc_bclink_name)) {
2562 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2563 (tipc_bclink_set_queue_limits(new_value) == 0))
2564 return tipc_cfg_reply_none();
2565 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2566 " (cannot change setting on broadcast link)");
2569 read_lock_bh(&tipc_net_lock);
2570 res = link_cmd_set_value(args->name, new_value, cmd);
2571 read_unlock_bh(&tipc_net_lock);
2573 return tipc_cfg_reply_error_string("cannot change link setting");
2575 return tipc_cfg_reply_none();
2579 * link_reset_statistics - reset link statistics
2580 * @l_ptr: pointer to link
2582 static void link_reset_statistics(struct tipc_link *l_ptr)
2584 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2585 l_ptr->stats.sent_info = l_ptr->next_out_no;
2586 l_ptr->stats.recv_info = l_ptr->next_in_no;
2589 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2592 struct tipc_link *l_ptr;
2593 struct tipc_node *node;
2594 unsigned int bearer_id;
2596 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2597 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2599 link_name = (char *)TLV_DATA(req_tlv_area);
2600 if (!strcmp(link_name, tipc_bclink_name)) {
2601 if (tipc_bclink_reset_stats())
2602 return tipc_cfg_reply_error_string("link not found");
2603 return tipc_cfg_reply_none();
2605 read_lock_bh(&tipc_net_lock);
2606 node = tipc_link_find_owner(link_name, &bearer_id);
2608 read_unlock_bh(&tipc_net_lock);
2609 return tipc_cfg_reply_error_string("link not found");
2611 tipc_node_lock(node);
2612 l_ptr = node->links[bearer_id];
2614 tipc_node_unlock(node);
2615 read_unlock_bh(&tipc_net_lock);
2616 return tipc_cfg_reply_error_string("link not found");
2618 link_reset_statistics(l_ptr);
2619 tipc_node_unlock(node);
2620 read_unlock_bh(&tipc_net_lock);
2621 return tipc_cfg_reply_none();
2625 * percent - convert count to a percentage of total (rounding up or down)
2627 static u32 percent(u32 count, u32 total)
2629 return (count * 100 + (total / 2)) / total;
2633 * tipc_link_stats - print link statistics
2635 * @buf: print buffer area
2636 * @buf_size: size of print buffer area
2638 * Returns length of print buffer data string (or 0 if error)
2640 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2642 struct tipc_link *l;
2643 struct tipc_stats *s;
2644 struct tipc_node *node;
2646 u32 profile_total = 0;
2647 unsigned int bearer_id;
2650 if (!strcmp(name, tipc_bclink_name))
2651 return tipc_bclink_stats(buf, buf_size);
2653 read_lock_bh(&tipc_net_lock);
2654 node = tipc_link_find_owner(name, &bearer_id);
2656 read_unlock_bh(&tipc_net_lock);
2659 tipc_node_lock(node);
2661 l = node->links[bearer_id];
2663 tipc_node_unlock(node);
2664 read_unlock_bh(&tipc_net_lock);
2670 if (tipc_link_is_active(l))
2672 else if (tipc_link_is_up(l))
2677 ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2678 " %s MTU:%u Priority:%u Tolerance:%u ms"
2679 " Window:%u packets\n",
2680 l->name, status, l->max_pkt, l->priority,
2681 l->tolerance, l->queue_limit[0]);
2683 ret += tipc_snprintf(buf + ret, buf_size - ret,
2684 " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2685 l->next_in_no - s->recv_info, s->recv_fragments,
2686 s->recv_fragmented, s->recv_bundles,
2689 ret += tipc_snprintf(buf + ret, buf_size - ret,
2690 " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2691 l->next_out_no - s->sent_info, s->sent_fragments,
2692 s->sent_fragmented, s->sent_bundles,
2695 profile_total = s->msg_length_counts;
2699 ret += tipc_snprintf(buf + ret, buf_size - ret,
2700 " TX profile sample:%u packets average:%u octets\n"
2701 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2702 "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2703 s->msg_length_counts,
2704 s->msg_lengths_total / profile_total,
2705 percent(s->msg_length_profile[0], profile_total),
2706 percent(s->msg_length_profile[1], profile_total),
2707 percent(s->msg_length_profile[2], profile_total),
2708 percent(s->msg_length_profile[3], profile_total),
2709 percent(s->msg_length_profile[4], profile_total),
2710 percent(s->msg_length_profile[5], profile_total),
2711 percent(s->msg_length_profile[6], profile_total));
2713 ret += tipc_snprintf(buf + ret, buf_size - ret,
2714 " RX states:%u probes:%u naks:%u defs:%u"
2715 " dups:%u\n", s->recv_states, s->recv_probes,
2716 s->recv_nacks, s->deferred_recv, s->duplicates);
2718 ret += tipc_snprintf(buf + ret, buf_size - ret,
2719 " TX states:%u probes:%u naks:%u acks:%u"
2720 " dups:%u\n", s->sent_states, s->sent_probes,
2721 s->sent_nacks, s->sent_acks, s->retransmitted);
2723 ret += tipc_snprintf(buf + ret, buf_size - ret,
2724 " Congestion link:%u Send queue"
2725 " max:%u avg:%u\n", s->link_congs,
2726 s->max_queue_sz, s->queue_sz_counts ?
2727 (s->accu_queue_sz / s->queue_sz_counts) : 0);
2729 tipc_node_unlock(node);
2730 read_unlock_bh(&tipc_net_lock);
2734 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2736 struct sk_buff *buf;
2737 struct tlv_desc *rep_tlv;
2742 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2743 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2745 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2749 rep_tlv = (struct tlv_desc *)buf->data;
2750 pb = TLV_DATA(rep_tlv);
2751 pb_len = ULTRA_STRING_MAX_LEN;
2752 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2756 return tipc_cfg_reply_error_string("link not found");
2758 str_len += 1; /* for "\0" */
2759 skb_put(buf, TLV_SPACE(str_len));
2760 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2766 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2767 * @dest: network address of destination node
2768 * @selector: used to select from set of active links
2770 * If no active link can be found, uses default maximum packet size.
2772 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2774 struct tipc_node *n_ptr;
2775 struct tipc_link *l_ptr;
2776 u32 res = MAX_PKT_DEFAULT;
2778 if (dest == tipc_own_addr)
2779 return MAX_MSG_SIZE;
2781 read_lock_bh(&tipc_net_lock);
2782 n_ptr = tipc_node_find(dest);
2784 tipc_node_lock(n_ptr);
2785 l_ptr = n_ptr->active_links[selector & 1];
2787 res = l_ptr->max_pkt;
2788 tipc_node_unlock(n_ptr);
2790 read_unlock_bh(&tipc_net_lock);
2794 static void link_print(struct tipc_link *l_ptr, const char *str)
2796 pr_info("%s Link %x<%s>:", str, l_ptr->addr, l_ptr->b_ptr->name);
2798 if (link_working_unknown(l_ptr))
2800 else if (link_reset_reset(l_ptr))
2802 else if (link_reset_unknown(l_ptr))
2804 else if (link_working_working(l_ptr))