2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, Ericsson AB
5 * Copyright (c) 2004-2007, 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"
46 * Out-of-range value for link session numbers
49 #define INVALID_SESSION 0x10000
55 #define STARTING_EVT 856384768 /* link processing trigger */
56 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
57 #define TIMEOUT_EVT 560817u /* link timer expired */
60 * The following two 'message types' is really just implementation
61 * data conveniently stored in the message header.
62 * They must not be considered part of the protocol
68 * State value stored in 'exp_msg_count'
71 #define START_CHANGEOVER 100000u
74 * struct link_name - deconstructed link name
75 * @addr_local: network address of node at this end
76 * @if_local: name of interface at this end
77 * @addr_peer: network address of node at far end
78 * @if_peer: name of interface at far end
83 char if_local[TIPC_MAX_IF_NAME];
85 char if_peer[TIPC_MAX_IF_NAME];
88 static void link_handle_out_of_seq_msg(struct link *l_ptr,
90 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
91 static int link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
92 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
93 static int link_send_sections_long(struct port *sender,
94 struct iovec const *msg_sect,
95 u32 num_sect, u32 destnode);
96 static void link_check_defragm_bufs(struct link *l_ptr);
97 static void link_state_event(struct link *l_ptr, u32 event);
98 static void link_reset_statistics(struct link *l_ptr);
99 static void link_print(struct link *l_ptr, const char *str);
100 static void link_start(struct link *l_ptr);
101 static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf);
104 * Simple link routines
107 static unsigned int align(unsigned int i)
109 return (i + 3) & ~3u;
112 static void link_init_max_pkt(struct link *l_ptr)
116 max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
117 if (max_pkt > MAX_MSG_SIZE)
118 max_pkt = MAX_MSG_SIZE;
120 l_ptr->max_pkt_target = max_pkt;
121 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
122 l_ptr->max_pkt = l_ptr->max_pkt_target;
124 l_ptr->max_pkt = MAX_PKT_DEFAULT;
126 l_ptr->max_pkt_probes = 0;
129 static u32 link_next_sent(struct link *l_ptr)
132 return msg_seqno(buf_msg(l_ptr->next_out));
133 return mod(l_ptr->next_out_no);
136 static u32 link_last_sent(struct link *l_ptr)
138 return mod(link_next_sent(l_ptr) - 1);
142 * Simple non-static link routines (i.e. referenced outside this file)
145 int tipc_link_is_up(struct link *l_ptr)
149 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
152 int tipc_link_is_active(struct link *l_ptr)
154 return (l_ptr->owner->active_links[0] == l_ptr) ||
155 (l_ptr->owner->active_links[1] == l_ptr);
159 * link_name_validate - validate & (optionally) deconstruct link name
160 * @name - ptr to link name string
161 * @name_parts - ptr to area for link name components (or NULL if not needed)
163 * Returns 1 if link name is valid, otherwise 0.
166 static int link_name_validate(const char *name, struct link_name *name_parts)
168 char name_copy[TIPC_MAX_LINK_NAME];
174 u32 z_local, c_local, n_local;
175 u32 z_peer, c_peer, n_peer;
179 /* copy link name & ensure length is OK */
181 name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
182 /* need above in case non-Posix strncpy() doesn't pad with nulls */
183 strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
184 if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
187 /* ensure all component parts of link name are present */
189 addr_local = name_copy;
190 if ((if_local = strchr(addr_local, ':')) == NULL)
193 if ((addr_peer = strchr(if_local, '-')) == NULL)
196 if_local_len = addr_peer - if_local;
197 if ((if_peer = strchr(addr_peer, ':')) == NULL)
200 if_peer_len = strlen(if_peer) + 1;
202 /* validate component parts of link name */
204 if ((sscanf(addr_local, "%u.%u.%u%c",
205 &z_local, &c_local, &n_local, &dummy) != 3) ||
206 (sscanf(addr_peer, "%u.%u.%u%c",
207 &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
208 (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
209 (z_peer > 255) || (c_peer > 4095) || (n_peer > 4095) ||
210 (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
211 (if_peer_len <= 1) || (if_peer_len > TIPC_MAX_IF_NAME) ||
212 (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
213 (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
216 /* return link name components, if necessary */
219 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
220 strcpy(name_parts->if_local, if_local);
221 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
222 strcpy(name_parts->if_peer, if_peer);
228 * link_timeout - handle expiration of link timer
229 * @l_ptr: pointer to link
231 * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
232 * with tipc_link_delete(). (There is no risk that the node will be deleted by
233 * another thread because tipc_link_delete() always cancels the link timer before
234 * tipc_node_delete() is called.)
237 static void link_timeout(struct link *l_ptr)
239 tipc_node_lock(l_ptr->owner);
241 /* update counters used in statistical profiling of send traffic */
243 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
244 l_ptr->stats.queue_sz_counts++;
246 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
247 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
249 if (l_ptr->first_out) {
250 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
251 u32 length = msg_size(msg);
253 if ((msg_user(msg) == MSG_FRAGMENTER) &&
254 (msg_type(msg) == FIRST_FRAGMENT)) {
255 length = msg_size(msg_get_wrapped(msg));
258 l_ptr->stats.msg_lengths_total += length;
259 l_ptr->stats.msg_length_counts++;
261 l_ptr->stats.msg_length_profile[0]++;
262 else if (length <= 256)
263 l_ptr->stats.msg_length_profile[1]++;
264 else if (length <= 1024)
265 l_ptr->stats.msg_length_profile[2]++;
266 else if (length <= 4096)
267 l_ptr->stats.msg_length_profile[3]++;
268 else if (length <= 16384)
269 l_ptr->stats.msg_length_profile[4]++;
270 else if (length <= 32768)
271 l_ptr->stats.msg_length_profile[5]++;
273 l_ptr->stats.msg_length_profile[6]++;
277 /* do all other link processing performed on a periodic basis */
279 link_check_defragm_bufs(l_ptr);
281 link_state_event(l_ptr, TIMEOUT_EVT);
284 tipc_link_push_queue(l_ptr);
286 tipc_node_unlock(l_ptr->owner);
289 static void link_set_timer(struct link *l_ptr, u32 time)
291 k_start_timer(&l_ptr->timer, time);
295 * tipc_link_create - create a new link
296 * @b_ptr: pointer to associated bearer
297 * @peer: network address of node at other end of link
298 * @media_addr: media address to use when sending messages over link
300 * Returns pointer to link.
303 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
304 const struct tipc_media_addr *media_addr)
307 struct tipc_msg *msg;
310 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
312 warn("Link creation failed, no memory\n");
317 if_name = strchr(b_ptr->publ.name, ':') + 1;
318 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
319 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
320 tipc_node(tipc_own_addr),
322 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
323 /* note: peer i/f is appended to link name by reset/activate */
324 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
325 l_ptr->checkpoint = 1;
326 l_ptr->b_ptr = b_ptr;
327 link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
328 l_ptr->state = RESET_UNKNOWN;
330 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
332 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
333 msg_set_size(msg, sizeof(l_ptr->proto_msg));
334 msg_set_session(msg, (tipc_random & 0xffff));
335 msg_set_bearer_id(msg, b_ptr->identity);
336 strcpy((char *)msg_data(msg), if_name);
338 l_ptr->priority = b_ptr->priority;
339 tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
341 link_init_max_pkt(l_ptr);
343 l_ptr->next_out_no = 1;
344 INIT_LIST_HEAD(&l_ptr->waiting_ports);
346 link_reset_statistics(l_ptr);
348 l_ptr->owner = tipc_node_attach_link(l_ptr);
354 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
355 list_add_tail(&l_ptr->link_list, &b_ptr->links);
356 tipc_k_signal((Handler)link_start, (unsigned long)l_ptr);
362 * tipc_link_delete - delete a link
363 * @l_ptr: pointer to link
365 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
366 * This routine must not grab the node lock until after link timer cancellation
367 * to avoid a potential deadlock situation.
370 void tipc_link_delete(struct link *l_ptr)
373 err("Attempt to delete non-existent link\n");
377 k_cancel_timer(&l_ptr->timer);
379 tipc_node_lock(l_ptr->owner);
380 tipc_link_reset(l_ptr);
381 tipc_node_detach_link(l_ptr->owner, l_ptr);
382 tipc_link_stop(l_ptr);
383 list_del_init(&l_ptr->link_list);
384 tipc_node_unlock(l_ptr->owner);
385 k_term_timer(&l_ptr->timer);
389 static void link_start(struct link *l_ptr)
391 link_state_event(l_ptr, STARTING_EVT);
395 * link_schedule_port - schedule port for deferred sending
396 * @l_ptr: pointer to link
397 * @origport: reference to sending port
398 * @sz: amount of data to be sent
400 * Schedules port for renewed sending of messages after link congestion
404 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
408 spin_lock_bh(&tipc_port_list_lock);
409 p_ptr = tipc_port_lock(origport);
413 if (!list_empty(&p_ptr->wait_list))
415 p_ptr->publ.congested = 1;
416 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
417 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
418 l_ptr->stats.link_congs++;
420 tipc_port_unlock(p_ptr);
422 spin_unlock_bh(&tipc_port_list_lock);
426 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
429 struct port *temp_p_ptr;
430 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
436 if (!spin_trylock_bh(&tipc_port_list_lock))
438 if (link_congested(l_ptr))
440 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
444 list_del_init(&p_ptr->wait_list);
445 spin_lock_bh(p_ptr->publ.lock);
446 p_ptr->publ.congested = 0;
447 p_ptr->wakeup(&p_ptr->publ);
448 win -= p_ptr->waiting_pkts;
449 spin_unlock_bh(p_ptr->publ.lock);
453 spin_unlock_bh(&tipc_port_list_lock);
457 * link_release_outqueue - purge link's outbound message queue
458 * @l_ptr: pointer to link
461 static void link_release_outqueue(struct link *l_ptr)
463 struct sk_buff *buf = l_ptr->first_out;
464 struct sk_buff *next;
471 l_ptr->first_out = NULL;
472 l_ptr->out_queue_size = 0;
476 * tipc_link_reset_fragments - purge link's inbound message fragments queue
477 * @l_ptr: pointer to link
480 void tipc_link_reset_fragments(struct link *l_ptr)
482 struct sk_buff *buf = l_ptr->defragm_buf;
483 struct sk_buff *next;
490 l_ptr->defragm_buf = NULL;
494 * tipc_link_stop - purge all inbound and outbound messages associated with link
495 * @l_ptr: pointer to link
498 void tipc_link_stop(struct link *l_ptr)
501 struct sk_buff *next;
503 buf = l_ptr->oldest_deferred_in;
510 buf = l_ptr->first_out;
517 tipc_link_reset_fragments(l_ptr);
519 buf_discard(l_ptr->proto_msg_queue);
520 l_ptr->proto_msg_queue = NULL;
523 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
524 #define link_send_event(fcn, l_ptr, up) do { } while (0)
526 void tipc_link_reset(struct link *l_ptr)
529 u32 prev_state = l_ptr->state;
530 u32 checkpoint = l_ptr->next_in_no;
531 int was_active_link = tipc_link_is_active(l_ptr);
533 msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
535 /* Link is down, accept any session */
536 l_ptr->peer_session = INVALID_SESSION;
538 /* Prepare for max packet size negotiation */
539 link_init_max_pkt(l_ptr);
541 l_ptr->state = RESET_UNKNOWN;
543 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
546 tipc_node_link_down(l_ptr->owner, l_ptr);
547 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
549 if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
550 l_ptr->owner->permit_changeover) {
551 l_ptr->reset_checkpoint = checkpoint;
552 l_ptr->exp_msg_count = START_CHANGEOVER;
555 /* Clean up all queues: */
557 link_release_outqueue(l_ptr);
558 buf_discard(l_ptr->proto_msg_queue);
559 l_ptr->proto_msg_queue = NULL;
560 buf = l_ptr->oldest_deferred_in;
562 struct sk_buff *next = buf->next;
566 if (!list_empty(&l_ptr->waiting_ports))
567 tipc_link_wakeup_ports(l_ptr, 1);
569 l_ptr->retransm_queue_head = 0;
570 l_ptr->retransm_queue_size = 0;
571 l_ptr->last_out = NULL;
572 l_ptr->first_out = NULL;
573 l_ptr->next_out = NULL;
574 l_ptr->unacked_window = 0;
575 l_ptr->checkpoint = 1;
576 l_ptr->next_out_no = 1;
577 l_ptr->deferred_inqueue_sz = 0;
578 l_ptr->oldest_deferred_in = NULL;
579 l_ptr->newest_deferred_in = NULL;
580 l_ptr->fsm_msg_cnt = 0;
581 l_ptr->stale_count = 0;
582 link_reset_statistics(l_ptr);
584 link_send_event(tipc_cfg_link_event, l_ptr, 0);
585 if (!in_own_cluster(l_ptr->addr))
586 link_send_event(tipc_disc_link_event, l_ptr, 0);
590 static void link_activate(struct link *l_ptr)
592 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
593 tipc_node_link_up(l_ptr->owner, l_ptr);
594 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
595 link_send_event(tipc_cfg_link_event, l_ptr, 1);
596 if (!in_own_cluster(l_ptr->addr))
597 link_send_event(tipc_disc_link_event, l_ptr, 1);
601 * link_state_event - link finite state machine
602 * @l_ptr: pointer to link
603 * @event: state machine event to process
606 static void link_state_event(struct link *l_ptr, unsigned event)
609 u32 cont_intv = l_ptr->continuity_interval;
611 if (!l_ptr->started && (event != STARTING_EVT))
612 return; /* Not yet. */
614 if (link_blocked(l_ptr)) {
615 if (event == TIMEOUT_EVT) {
616 link_set_timer(l_ptr, cont_intv);
618 return; /* Changeover going on */
621 switch (l_ptr->state) {
622 case WORKING_WORKING:
624 case TRAFFIC_MSG_EVT:
628 if (l_ptr->next_in_no != l_ptr->checkpoint) {
629 l_ptr->checkpoint = l_ptr->next_in_no;
630 if (tipc_bclink_acks_missing(l_ptr->owner)) {
631 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
633 l_ptr->fsm_msg_cnt++;
634 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
635 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
637 l_ptr->fsm_msg_cnt++;
639 link_set_timer(l_ptr, cont_intv);
642 l_ptr->state = WORKING_UNKNOWN;
643 l_ptr->fsm_msg_cnt = 0;
644 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
645 l_ptr->fsm_msg_cnt++;
646 link_set_timer(l_ptr, cont_intv / 4);
649 info("Resetting link <%s>, requested by peer\n",
651 tipc_link_reset(l_ptr);
652 l_ptr->state = RESET_RESET;
653 l_ptr->fsm_msg_cnt = 0;
654 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
655 l_ptr->fsm_msg_cnt++;
656 link_set_timer(l_ptr, cont_intv);
659 err("Unknown link event %u in WW state\n", event);
662 case WORKING_UNKNOWN:
664 case TRAFFIC_MSG_EVT:
666 l_ptr->state = WORKING_WORKING;
667 l_ptr->fsm_msg_cnt = 0;
668 link_set_timer(l_ptr, cont_intv);
671 info("Resetting link <%s>, requested by peer "
672 "while probing\n", l_ptr->name);
673 tipc_link_reset(l_ptr);
674 l_ptr->state = RESET_RESET;
675 l_ptr->fsm_msg_cnt = 0;
676 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
677 l_ptr->fsm_msg_cnt++;
678 link_set_timer(l_ptr, cont_intv);
681 if (l_ptr->next_in_no != l_ptr->checkpoint) {
682 l_ptr->state = WORKING_WORKING;
683 l_ptr->fsm_msg_cnt = 0;
684 l_ptr->checkpoint = l_ptr->next_in_no;
685 if (tipc_bclink_acks_missing(l_ptr->owner)) {
686 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
688 l_ptr->fsm_msg_cnt++;
690 link_set_timer(l_ptr, cont_intv);
691 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
692 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
694 l_ptr->fsm_msg_cnt++;
695 link_set_timer(l_ptr, cont_intv / 4);
696 } else { /* Link has failed */
697 warn("Resetting link <%s>, peer not responding\n",
699 tipc_link_reset(l_ptr);
700 l_ptr->state = RESET_UNKNOWN;
701 l_ptr->fsm_msg_cnt = 0;
702 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
704 l_ptr->fsm_msg_cnt++;
705 link_set_timer(l_ptr, cont_intv);
709 err("Unknown link event %u in WU state\n", event);
714 case TRAFFIC_MSG_EVT:
717 other = l_ptr->owner->active_links[0];
718 if (other && link_working_unknown(other))
720 l_ptr->state = WORKING_WORKING;
721 l_ptr->fsm_msg_cnt = 0;
722 link_activate(l_ptr);
723 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
724 l_ptr->fsm_msg_cnt++;
725 link_set_timer(l_ptr, cont_intv);
728 l_ptr->state = RESET_RESET;
729 l_ptr->fsm_msg_cnt = 0;
730 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
731 l_ptr->fsm_msg_cnt++;
732 link_set_timer(l_ptr, cont_intv);
738 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
739 l_ptr->fsm_msg_cnt++;
740 link_set_timer(l_ptr, cont_intv);
743 err("Unknown link event %u in RU state\n", event);
748 case TRAFFIC_MSG_EVT:
750 other = l_ptr->owner->active_links[0];
751 if (other && link_working_unknown(other))
753 l_ptr->state = WORKING_WORKING;
754 l_ptr->fsm_msg_cnt = 0;
755 link_activate(l_ptr);
756 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
757 l_ptr->fsm_msg_cnt++;
758 link_set_timer(l_ptr, cont_intv);
763 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
764 l_ptr->fsm_msg_cnt++;
765 link_set_timer(l_ptr, cont_intv);
768 err("Unknown link event %u in RR state\n", event);
772 err("Unknown link state %u/%u\n", l_ptr->state, event);
777 * link_bundle_buf(): Append contents of a buffer to
778 * the tail of an existing one.
781 static int link_bundle_buf(struct link *l_ptr,
782 struct sk_buff *bundler,
785 struct tipc_msg *bundler_msg = buf_msg(bundler);
786 struct tipc_msg *msg = buf_msg(buf);
787 u32 size = msg_size(msg);
788 u32 bundle_size = msg_size(bundler_msg);
789 u32 to_pos = align(bundle_size);
790 u32 pad = to_pos - bundle_size;
792 if (msg_user(bundler_msg) != MSG_BUNDLER)
794 if (msg_type(bundler_msg) != OPEN_MSG)
796 if (skb_tailroom(bundler) < (pad + size))
798 if (l_ptr->max_pkt < (to_pos + size))
801 skb_put(bundler, pad + size);
802 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
803 msg_set_size(bundler_msg, to_pos + size);
804 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
806 l_ptr->stats.sent_bundled++;
810 static void link_add_to_outqueue(struct link *l_ptr,
812 struct tipc_msg *msg)
814 u32 ack = mod(l_ptr->next_in_no - 1);
815 u32 seqno = mod(l_ptr->next_out_no++);
817 msg_set_word(msg, 2, ((ack << 16) | seqno));
818 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
820 if (l_ptr->first_out) {
821 l_ptr->last_out->next = buf;
822 l_ptr->last_out = buf;
824 l_ptr->first_out = l_ptr->last_out = buf;
825 l_ptr->out_queue_size++;
829 * tipc_link_send_buf() is the 'full path' for messages, called from
830 * inside TIPC when the 'fast path' in tipc_send_buf
831 * has failed, and from link_send()
834 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
836 struct tipc_msg *msg = buf_msg(buf);
837 u32 size = msg_size(msg);
838 u32 dsz = msg_data_sz(msg);
839 u32 queue_size = l_ptr->out_queue_size;
840 u32 imp = tipc_msg_tot_importance(msg);
841 u32 queue_limit = l_ptr->queue_limit[imp];
842 u32 max_packet = l_ptr->max_pkt;
844 msg_set_prevnode(msg, tipc_own_addr); /* If routed message */
846 /* Match msg importance against queue limits: */
848 if (unlikely(queue_size >= queue_limit)) {
849 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
850 return link_schedule_port(l_ptr, msg_origport(msg),
854 if (imp > CONN_MANAGER) {
855 warn("Resetting link <%s>, send queue full", l_ptr->name);
856 tipc_link_reset(l_ptr);
861 /* Fragmentation needed ? */
863 if (size > max_packet)
864 return link_send_long_buf(l_ptr, buf);
866 /* Packet can be queued or sent: */
868 if (queue_size > l_ptr->stats.max_queue_sz)
869 l_ptr->stats.max_queue_sz = queue_size;
871 if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
872 !link_congested(l_ptr))) {
873 link_add_to_outqueue(l_ptr, buf, msg);
875 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
876 l_ptr->unacked_window = 0;
878 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
879 l_ptr->stats.bearer_congs++;
880 l_ptr->next_out = buf;
884 /* Congestion: can message be bundled ?: */
886 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
887 (msg_user(msg) != MSG_FRAGMENTER)) {
889 /* Try adding message to an existing bundle */
891 if (l_ptr->next_out &&
892 link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
893 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
897 /* Try creating a new bundle */
899 if (size <= max_packet * 2 / 3) {
900 struct sk_buff *bundler = tipc_buf_acquire(max_packet);
901 struct tipc_msg bundler_hdr;
904 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
905 INT_H_SIZE, l_ptr->addr);
906 skb_copy_to_linear_data(bundler, &bundler_hdr,
908 skb_trim(bundler, INT_H_SIZE);
909 link_bundle_buf(l_ptr, bundler, buf);
912 l_ptr->stats.sent_bundles++;
916 if (!l_ptr->next_out)
917 l_ptr->next_out = buf;
918 link_add_to_outqueue(l_ptr, buf, msg);
919 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
924 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
925 * not been selected yet, and the the owner node is not locked
926 * Called by TIPC internal users, e.g. the name distributor
929 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
932 struct tipc_node *n_ptr;
933 int res = -ELINKCONG;
935 read_lock_bh(&tipc_net_lock);
936 n_ptr = tipc_node_find(dest);
938 tipc_node_lock(n_ptr);
939 l_ptr = n_ptr->active_links[selector & 1];
941 res = tipc_link_send_buf(l_ptr, buf);
945 tipc_node_unlock(n_ptr);
949 read_unlock_bh(&tipc_net_lock);
954 * link_send_buf_fast: Entry for data messages where the
955 * destination link is known and the header is complete,
956 * inclusive total message length. Very time critical.
957 * Link is locked. Returns user data length.
960 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
963 struct tipc_msg *msg = buf_msg(buf);
964 int res = msg_data_sz(msg);
966 if (likely(!link_congested(l_ptr))) {
967 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
968 if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
969 link_add_to_outqueue(l_ptr, buf, msg);
970 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
971 &l_ptr->media_addr))) {
972 l_ptr->unacked_window = 0;
975 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
976 l_ptr->stats.bearer_congs++;
977 l_ptr->next_out = buf;
981 *used_max_pkt = l_ptr->max_pkt;
983 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
987 * tipc_send_buf_fast: Entry for data messages where the
988 * destination node is known and the header is complete,
989 * inclusive total message length.
990 * Returns user data length.
992 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
995 struct tipc_node *n_ptr;
997 u32 selector = msg_origport(buf_msg(buf)) & 1;
1000 if (destnode == tipc_own_addr)
1001 return tipc_port_recv_msg(buf);
1003 read_lock_bh(&tipc_net_lock);
1004 n_ptr = tipc_node_find(destnode);
1005 if (likely(n_ptr)) {
1006 tipc_node_lock(n_ptr);
1007 l_ptr = n_ptr->active_links[selector];
1008 if (likely(l_ptr)) {
1009 res = link_send_buf_fast(l_ptr, buf, &dummy);
1010 tipc_node_unlock(n_ptr);
1011 read_unlock_bh(&tipc_net_lock);
1014 tipc_node_unlock(n_ptr);
1016 read_unlock_bh(&tipc_net_lock);
1017 res = msg_data_sz(buf_msg(buf));
1018 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1024 * tipc_link_send_sections_fast: Entry for messages where the
1025 * destination processor is known and the header is complete,
1026 * except for total message length.
1027 * Returns user data length or errno.
1029 int tipc_link_send_sections_fast(struct port *sender,
1030 struct iovec const *msg_sect,
1034 struct tipc_msg *hdr = &sender->publ.phdr;
1036 struct sk_buff *buf;
1037 struct tipc_node *node;
1039 u32 selector = msg_origport(hdr) & 1;
1043 * Try building message using port's max_pkt hint.
1044 * (Must not hold any locks while building message.)
1047 res = tipc_msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1048 !sender->user_port, &buf);
1050 read_lock_bh(&tipc_net_lock);
1051 node = tipc_node_find(destaddr);
1053 tipc_node_lock(node);
1054 l_ptr = node->active_links[selector];
1055 if (likely(l_ptr)) {
1057 res = link_send_buf_fast(l_ptr, buf,
1058 &sender->publ.max_pkt);
1059 if (unlikely(res < 0))
1062 tipc_node_unlock(node);
1063 read_unlock_bh(&tipc_net_lock);
1067 /* Exit if build request was invalid */
1069 if (unlikely(res < 0))
1072 /* Exit if link (or bearer) is congested */
1074 if (link_congested(l_ptr) ||
1075 !list_empty(&l_ptr->b_ptr->cong_links)) {
1076 res = link_schedule_port(l_ptr,
1077 sender->publ.ref, res);
1082 * Message size exceeds max_pkt hint; update hint,
1083 * then re-try fast path or fragment the message
1086 sender->publ.max_pkt = l_ptr->max_pkt;
1087 tipc_node_unlock(node);
1088 read_unlock_bh(&tipc_net_lock);
1091 if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1094 return link_send_sections_long(sender, msg_sect,
1095 num_sect, destaddr);
1097 tipc_node_unlock(node);
1099 read_unlock_bh(&tipc_net_lock);
1101 /* Couldn't find a link to the destination node */
1104 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1106 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1112 * link_send_sections_long(): Entry for long messages where the
1113 * destination node is known and the header is complete,
1114 * inclusive total message length.
1115 * Link and bearer congestion status have been checked to be ok,
1116 * and are ignored if they change.
1118 * Note that fragments do not use the full link MTU so that they won't have
1119 * to undergo refragmentation if link changeover causes them to be sent
1120 * over another link with an additional tunnel header added as prefix.
1121 * (Refragmentation will still occur if the other link has a smaller MTU.)
1123 * Returns user data length or errno.
1125 static int link_send_sections_long(struct port *sender,
1126 struct iovec const *msg_sect,
1131 struct tipc_node *node;
1132 struct tipc_msg *hdr = &sender->publ.phdr;
1133 u32 dsz = msg_data_sz(hdr);
1134 u32 max_pkt, fragm_sz, rest;
1135 struct tipc_msg fragm_hdr;
1136 struct sk_buff *buf, *buf_chain, *prev;
1137 u32 fragm_crs, fragm_rest, hsz, sect_rest;
1138 const unchar *sect_crs;
1144 max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1145 /* leave room for tunnel header in case of link changeover */
1146 fragm_sz = max_pkt - INT_H_SIZE;
1147 /* leave room for fragmentation header in each fragment */
1155 /* Prepare reusable fragment header: */
1157 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1158 INT_H_SIZE, msg_destnode(hdr));
1159 msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1160 msg_set_size(&fragm_hdr, max_pkt);
1161 msg_set_fragm_no(&fragm_hdr, 1);
1163 /* Prepare header of first fragment: */
1165 buf_chain = buf = tipc_buf_acquire(max_pkt);
1169 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1170 hsz = msg_hdr_sz(hdr);
1171 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1173 /* Chop up message: */
1175 fragm_crs = INT_H_SIZE + hsz;
1176 fragm_rest = fragm_sz - hsz;
1178 do { /* For all sections */
1182 sect_rest = msg_sect[++curr_sect].iov_len;
1183 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1186 if (sect_rest < fragm_rest)
1191 if (likely(!sender->user_port)) {
1192 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1194 for (; buf_chain; buf_chain = buf) {
1195 buf = buf_chain->next;
1196 buf_discard(buf_chain);
1201 skb_copy_to_linear_data_offset(buf, fragm_crs,
1209 if (!fragm_rest && rest) {
1211 /* Initiate new fragment: */
1212 if (rest <= fragm_sz) {
1214 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1216 msg_set_type(&fragm_hdr, FRAGMENT);
1218 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1219 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1221 buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1227 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1228 fragm_crs = INT_H_SIZE;
1229 fragm_rest = fragm_sz;
1234 * Now we have a buffer chain. Select a link and check
1235 * that packet size is still OK
1237 node = tipc_node_find(destaddr);
1239 tipc_node_lock(node);
1240 l_ptr = node->active_links[sender->publ.ref & 1];
1242 tipc_node_unlock(node);
1245 if (l_ptr->max_pkt < max_pkt) {
1246 sender->publ.max_pkt = l_ptr->max_pkt;
1247 tipc_node_unlock(node);
1248 for (; buf_chain; buf_chain = buf) {
1249 buf = buf_chain->next;
1250 buf_discard(buf_chain);
1256 for (; buf_chain; buf_chain = buf) {
1257 buf = buf_chain->next;
1258 buf_discard(buf_chain);
1260 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1264 /* Append whole chain to send queue: */
1267 l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1268 if (!l_ptr->next_out)
1269 l_ptr->next_out = buf_chain;
1270 l_ptr->stats.sent_fragmented++;
1272 struct sk_buff *next = buf->next;
1273 struct tipc_msg *msg = buf_msg(buf);
1275 l_ptr->stats.sent_fragments++;
1276 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1277 link_add_to_outqueue(l_ptr, buf, msg);
1281 /* Send it, if possible: */
1283 tipc_link_push_queue(l_ptr);
1284 tipc_node_unlock(node);
1289 * tipc_link_push_packet: Push one unsent packet to the media
1291 u32 tipc_link_push_packet(struct link *l_ptr)
1293 struct sk_buff *buf = l_ptr->first_out;
1294 u32 r_q_size = l_ptr->retransm_queue_size;
1295 u32 r_q_head = l_ptr->retransm_queue_head;
1297 /* Step to position where retransmission failed, if any, */
1298 /* consider that buffers may have been released in meantime */
1300 if (r_q_size && buf) {
1301 u32 last = lesser(mod(r_q_head + r_q_size),
1302 link_last_sent(l_ptr));
1303 u32 first = msg_seqno(buf_msg(buf));
1305 while (buf && less(first, r_q_head)) {
1306 first = mod(first + 1);
1309 l_ptr->retransm_queue_head = r_q_head = first;
1310 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1313 /* Continue retransmission now, if there is anything: */
1315 if (r_q_size && buf) {
1316 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1317 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1318 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1319 l_ptr->retransm_queue_head = mod(++r_q_head);
1320 l_ptr->retransm_queue_size = --r_q_size;
1321 l_ptr->stats.retransmitted++;
1324 l_ptr->stats.bearer_congs++;
1329 /* Send deferred protocol message, if any: */
1331 buf = l_ptr->proto_msg_queue;
1333 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1334 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1335 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1336 l_ptr->unacked_window = 0;
1338 l_ptr->proto_msg_queue = NULL;
1341 l_ptr->stats.bearer_congs++;
1346 /* Send one deferred data message, if send window not full: */
1348 buf = l_ptr->next_out;
1350 struct tipc_msg *msg = buf_msg(buf);
1351 u32 next = msg_seqno(msg);
1352 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1354 if (mod(next - first) < l_ptr->queue_limit[0]) {
1355 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1356 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1357 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1358 if (msg_user(msg) == MSG_BUNDLER)
1359 msg_set_type(msg, CLOSED_MSG);
1360 l_ptr->next_out = buf->next;
1363 l_ptr->stats.bearer_congs++;
1368 return PUSH_FINISHED;
1372 * push_queue(): push out the unsent messages of a link where
1373 * congestion has abated. Node is locked
1375 void tipc_link_push_queue(struct link *l_ptr)
1379 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1383 res = tipc_link_push_packet(l_ptr);
1386 if (res == PUSH_FAILED)
1387 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1390 static void link_reset_all(unsigned long addr)
1392 struct tipc_node *n_ptr;
1393 char addr_string[16];
1396 read_lock_bh(&tipc_net_lock);
1397 n_ptr = tipc_node_find((u32)addr);
1399 read_unlock_bh(&tipc_net_lock);
1400 return; /* node no longer exists */
1403 tipc_node_lock(n_ptr);
1405 warn("Resetting all links to %s\n",
1406 tipc_addr_string_fill(addr_string, n_ptr->addr));
1408 for (i = 0; i < MAX_BEARERS; i++) {
1409 if (n_ptr->links[i]) {
1410 link_print(n_ptr->links[i], "Resetting link\n");
1411 tipc_link_reset(n_ptr->links[i]);
1415 tipc_node_unlock(n_ptr);
1416 read_unlock_bh(&tipc_net_lock);
1419 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1421 struct tipc_msg *msg = buf_msg(buf);
1423 warn("Retransmission failure on link <%s>\n", l_ptr->name);
1427 /* Handle failure on standard link */
1429 link_print(l_ptr, "Resetting link\n");
1430 tipc_link_reset(l_ptr);
1434 /* Handle failure on broadcast link */
1436 struct tipc_node *n_ptr;
1437 char addr_string[16];
1439 info("Msg seq number: %u, ", msg_seqno(msg));
1440 info("Outstanding acks: %lu\n",
1441 (unsigned long) TIPC_SKB_CB(buf)->handle);
1443 n_ptr = l_ptr->owner->next;
1444 tipc_node_lock(n_ptr);
1446 tipc_addr_string_fill(addr_string, n_ptr->addr);
1447 info("Multicast link info for %s\n", addr_string);
1448 info("Supported: %d, ", n_ptr->bclink.supported);
1449 info("Acked: %u\n", n_ptr->bclink.acked);
1450 info("Last in: %u, ", n_ptr->bclink.last_in);
1451 info("Gap after: %u, ", n_ptr->bclink.gap_after);
1452 info("Gap to: %u\n", n_ptr->bclink.gap_to);
1453 info("Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1455 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1457 tipc_node_unlock(n_ptr);
1459 l_ptr->stale_count = 0;
1463 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1466 struct tipc_msg *msg;
1473 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1474 if (l_ptr->retransm_queue_size == 0) {
1475 l_ptr->retransm_queue_head = msg_seqno(msg);
1476 l_ptr->retransm_queue_size = retransmits;
1478 err("Unexpected retransmit on link %s (qsize=%d)\n",
1479 l_ptr->name, l_ptr->retransm_queue_size);
1483 /* Detect repeated retransmit failures on uncongested bearer */
1485 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1486 if (++l_ptr->stale_count > 100) {
1487 link_retransmit_failure(l_ptr, buf);
1491 l_ptr->last_retransmitted = msg_seqno(msg);
1492 l_ptr->stale_count = 1;
1496 while (retransmits && (buf != l_ptr->next_out) && buf) {
1498 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1499 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1500 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1503 l_ptr->stats.retransmitted++;
1505 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1506 l_ptr->stats.bearer_congs++;
1507 l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1508 l_ptr->retransm_queue_size = retransmits;
1513 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1517 * link_insert_deferred_queue - insert deferred messages back into receive chain
1520 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1521 struct sk_buff *buf)
1525 if (l_ptr->oldest_deferred_in == NULL)
1528 seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1529 if (seq_no == mod(l_ptr->next_in_no)) {
1530 l_ptr->newest_deferred_in->next = buf;
1531 buf = l_ptr->oldest_deferred_in;
1532 l_ptr->oldest_deferred_in = NULL;
1533 l_ptr->deferred_inqueue_sz = 0;
1539 * link_recv_buf_validate - validate basic format of received message
1541 * This routine ensures a TIPC message has an acceptable header, and at least
1542 * as much data as the header indicates it should. The routine also ensures
1543 * that the entire message header is stored in the main fragment of the message
1544 * buffer, to simplify future access to message header fields.
1546 * Note: Having extra info present in the message header or data areas is OK.
1547 * TIPC will ignore the excess, under the assumption that it is optional info
1548 * introduced by a later release of the protocol.
1551 static int link_recv_buf_validate(struct sk_buff *buf)
1553 static u32 min_data_hdr_size[8] = {
1554 SHORT_H_SIZE, MCAST_H_SIZE, LONG_H_SIZE, DIR_MSG_H_SIZE,
1555 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1558 struct tipc_msg *msg;
1564 if (unlikely(buf->len < MIN_H_SIZE))
1567 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1571 if (unlikely(msg_version(msg) != TIPC_VERSION))
1574 size = msg_size(msg);
1575 hdr_size = msg_hdr_sz(msg);
1576 min_hdr_size = msg_isdata(msg) ?
1577 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1579 if (unlikely((hdr_size < min_hdr_size) ||
1580 (size < hdr_size) ||
1581 (buf->len < size) ||
1582 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1585 return pskb_may_pull(buf, hdr_size);
1589 * tipc_recv_msg - process TIPC messages arriving from off-node
1590 * @head: pointer to message buffer chain
1591 * @tb_ptr: pointer to bearer message arrived on
1593 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1594 * structure (i.e. cannot be NULL), but bearer can be inactive.
1597 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1599 read_lock_bh(&tipc_net_lock);
1601 struct bearer *b_ptr = (struct bearer *)tb_ptr;
1602 struct tipc_node *n_ptr;
1604 struct sk_buff *crs;
1605 struct sk_buff *buf = head;
1606 struct tipc_msg *msg;
1614 /* Ensure bearer is still enabled */
1616 if (unlikely(!b_ptr->active))
1619 /* Ensure message is well-formed */
1621 if (unlikely(!link_recv_buf_validate(buf)))
1624 /* Ensure message data is a single contiguous unit */
1626 if (unlikely(buf_linearize(buf))) {
1630 /* Handle arrival of a non-unicast link message */
1634 if (unlikely(msg_non_seq(msg))) {
1635 if (msg_user(msg) == LINK_CONFIG)
1636 tipc_disc_recv_msg(buf, b_ptr);
1638 tipc_bclink_recv_pkt(buf);
1642 if (unlikely(!msg_short(msg) &&
1643 (msg_destnode(msg) != tipc_own_addr)))
1646 /* Discard non-routeable messages destined for another node */
1648 if (unlikely(!msg_isdata(msg) &&
1649 (msg_destnode(msg) != tipc_own_addr))) {
1650 if ((msg_user(msg) != CONN_MANAGER) &&
1651 (msg_user(msg) != MSG_FRAGMENTER))
1655 /* Locate neighboring node that sent message */
1657 n_ptr = tipc_node_find(msg_prevnode(msg));
1658 if (unlikely(!n_ptr))
1660 tipc_node_lock(n_ptr);
1662 /* Don't talk to neighbor during cleanup after last session */
1664 if (n_ptr->cleanup_required) {
1665 tipc_node_unlock(n_ptr);
1669 /* Locate unicast link endpoint that should handle message */
1671 l_ptr = n_ptr->links[b_ptr->identity];
1672 if (unlikely(!l_ptr)) {
1673 tipc_node_unlock(n_ptr);
1677 /* Validate message sequence number info */
1679 seq_no = msg_seqno(msg);
1680 ackd = msg_ack(msg);
1682 /* Release acked messages */
1684 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1685 if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1686 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1689 crs = l_ptr->first_out;
1690 while ((crs != l_ptr->next_out) &&
1691 less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1692 struct sk_buff *next = crs->next;
1699 l_ptr->first_out = crs;
1700 l_ptr->out_queue_size -= released;
1703 /* Try sending any messages link endpoint has pending */
1705 if (unlikely(l_ptr->next_out))
1706 tipc_link_push_queue(l_ptr);
1707 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1708 tipc_link_wakeup_ports(l_ptr, 0);
1709 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1710 l_ptr->stats.sent_acks++;
1711 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1714 /* Now (finally!) process the incoming message */
1717 if (likely(link_working_working(l_ptr))) {
1718 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1719 l_ptr->next_in_no++;
1720 if (unlikely(l_ptr->oldest_deferred_in))
1721 head = link_insert_deferred_queue(l_ptr,
1723 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1725 if (likely(msg_isdata(msg))) {
1726 tipc_node_unlock(n_ptr);
1727 tipc_port_recv_msg(buf);
1730 switch (msg_user(msg)) {
1732 l_ptr->stats.recv_bundles++;
1733 l_ptr->stats.recv_bundled +=
1735 tipc_node_unlock(n_ptr);
1736 tipc_link_recv_bundle(buf);
1738 case ROUTE_DISTRIBUTOR:
1739 tipc_node_unlock(n_ptr);
1742 case NAME_DISTRIBUTOR:
1743 tipc_node_unlock(n_ptr);
1744 tipc_named_recv(buf);
1747 tipc_node_unlock(n_ptr);
1748 tipc_port_recv_proto_msg(buf);
1750 case MSG_FRAGMENTER:
1751 l_ptr->stats.recv_fragments++;
1752 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1754 l_ptr->stats.recv_fragmented++;
1758 case CHANGEOVER_PROTOCOL:
1759 type = msg_type(msg);
1760 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1762 seq_no = msg_seqno(msg);
1763 if (type == ORIGINAL_MSG)
1765 goto protocol_check;
1770 tipc_node_unlock(n_ptr);
1771 tipc_net_route_msg(buf);
1774 link_handle_out_of_seq_msg(l_ptr, buf);
1775 head = link_insert_deferred_queue(l_ptr, head);
1776 tipc_node_unlock(n_ptr);
1780 if (msg_user(msg) == LINK_PROTOCOL) {
1781 link_recv_proto_msg(l_ptr, buf);
1782 head = link_insert_deferred_queue(l_ptr, head);
1783 tipc_node_unlock(n_ptr);
1786 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1788 if (link_working_working(l_ptr)) {
1789 /* Re-insert in front of queue */
1792 tipc_node_unlock(n_ptr);
1795 tipc_node_unlock(n_ptr);
1799 read_unlock_bh(&tipc_net_lock);
1803 * link_defer_buf(): Sort a received out-of-sequence packet
1804 * into the deferred reception queue.
1805 * Returns the increase of the queue length,i.e. 0 or 1
1808 u32 tipc_link_defer_pkt(struct sk_buff **head,
1809 struct sk_buff **tail,
1810 struct sk_buff *buf)
1812 struct sk_buff *prev = NULL;
1813 struct sk_buff *crs = *head;
1814 u32 seq_no = msg_seqno(buf_msg(buf));
1819 if (*head == NULL) {
1820 *head = *tail = buf;
1825 if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1826 (*tail)->next = buf;
1831 /* Scan through queue and sort it in */
1833 struct tipc_msg *msg = buf_msg(crs);
1835 if (less(seq_no, msg_seqno(msg))) {
1843 if (seq_no == msg_seqno(msg)) {
1850 /* Message is a duplicate of an existing message */
1857 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1860 static void link_handle_out_of_seq_msg(struct link *l_ptr,
1861 struct sk_buff *buf)
1863 u32 seq_no = msg_seqno(buf_msg(buf));
1865 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1866 link_recv_proto_msg(l_ptr, buf);
1870 /* Record OOS packet arrival (force mismatch on next timeout) */
1872 l_ptr->checkpoint--;
1875 * Discard packet if a duplicate; otherwise add it to deferred queue
1876 * and notify peer of gap as per protocol specification
1879 if (less(seq_no, mod(l_ptr->next_in_no))) {
1880 l_ptr->stats.duplicates++;
1885 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1886 &l_ptr->newest_deferred_in, buf)) {
1887 l_ptr->deferred_inqueue_sz++;
1888 l_ptr->stats.deferred_recv++;
1889 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1890 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1892 l_ptr->stats.duplicates++;
1896 * Send protocol message to the other endpoint.
1898 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
1899 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1901 struct sk_buff *buf = NULL;
1902 struct tipc_msg *msg = l_ptr->pmsg;
1903 u32 msg_size = sizeof(l_ptr->proto_msg);
1905 if (link_blocked(l_ptr))
1907 msg_set_type(msg, msg_typ);
1908 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1909 msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
1910 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1912 if (msg_typ == STATE_MSG) {
1913 u32 next_sent = mod(l_ptr->next_out_no);
1915 if (!tipc_link_is_up(l_ptr))
1917 if (l_ptr->next_out)
1918 next_sent = msg_seqno(buf_msg(l_ptr->next_out));
1919 msg_set_next_sent(msg, next_sent);
1920 if (l_ptr->oldest_deferred_in) {
1921 u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1922 gap = mod(rec - mod(l_ptr->next_in_no));
1924 msg_set_seq_gap(msg, gap);
1926 l_ptr->stats.sent_nacks++;
1927 msg_set_link_tolerance(msg, tolerance);
1928 msg_set_linkprio(msg, priority);
1929 msg_set_max_pkt(msg, ack_mtu);
1930 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1931 msg_set_probe(msg, probe_msg != 0);
1933 u32 mtu = l_ptr->max_pkt;
1935 if ((mtu < l_ptr->max_pkt_target) &&
1936 link_working_working(l_ptr) &&
1937 l_ptr->fsm_msg_cnt) {
1938 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1939 if (l_ptr->max_pkt_probes == 10) {
1940 l_ptr->max_pkt_target = (msg_size - 4);
1941 l_ptr->max_pkt_probes = 0;
1942 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1944 l_ptr->max_pkt_probes++;
1947 l_ptr->stats.sent_probes++;
1949 l_ptr->stats.sent_states++;
1950 } else { /* RESET_MSG or ACTIVATE_MSG */
1951 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1952 msg_set_seq_gap(msg, 0);
1953 msg_set_next_sent(msg, 1);
1954 msg_set_link_tolerance(msg, l_ptr->tolerance);
1955 msg_set_linkprio(msg, l_ptr->priority);
1956 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1959 if (tipc_node_has_redundant_links(l_ptr->owner)) {
1960 msg_set_redundant_link(msg);
1962 msg_clear_redundant_link(msg);
1964 msg_set_linkprio(msg, l_ptr->priority);
1966 /* Ensure sequence number will not fit : */
1968 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1972 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1973 if (!l_ptr->proto_msg_queue) {
1974 l_ptr->proto_msg_queue =
1975 tipc_buf_acquire(sizeof(l_ptr->proto_msg));
1977 buf = l_ptr->proto_msg_queue;
1980 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1983 msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
1985 /* Message can be sent */
1987 buf = tipc_buf_acquire(msg_size);
1991 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1992 msg_set_size(buf_msg(buf), msg_size);
1994 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1995 l_ptr->unacked_window = 0;
2000 /* New congestion */
2001 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2002 l_ptr->proto_msg_queue = buf;
2003 l_ptr->stats.bearer_congs++;
2007 * Receive protocol message :
2008 * Note that network plane id propagates through the network, and may
2009 * change at any time. The node with lowest address rules
2012 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2018 struct tipc_msg *msg = buf_msg(buf);
2020 if (link_blocked(l_ptr))
2023 /* record unnumbered packet arrival (force mismatch on next timeout) */
2025 l_ptr->checkpoint--;
2027 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2028 if (tipc_own_addr > msg_prevnode(msg))
2029 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2031 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2033 switch (msg_type(msg)) {
2036 if (!link_working_unknown(l_ptr) &&
2037 (l_ptr->peer_session != INVALID_SESSION)) {
2038 if (msg_session(msg) == l_ptr->peer_session)
2039 break; /* duplicate: ignore */
2043 /* Update link settings according other endpoint's values */
2045 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2047 if ((msg_tol = msg_link_tolerance(msg)) &&
2048 (msg_tol > l_ptr->tolerance))
2049 link_set_supervision_props(l_ptr, msg_tol);
2051 if (msg_linkprio(msg) > l_ptr->priority)
2052 l_ptr->priority = msg_linkprio(msg);
2054 max_pkt_info = msg_max_pkt(msg);
2056 if (max_pkt_info < l_ptr->max_pkt_target)
2057 l_ptr->max_pkt_target = max_pkt_info;
2058 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2059 l_ptr->max_pkt = l_ptr->max_pkt_target;
2061 l_ptr->max_pkt = l_ptr->max_pkt_target;
2063 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2065 link_state_event(l_ptr, msg_type(msg));
2067 l_ptr->peer_session = msg_session(msg);
2068 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2070 /* Synchronize broadcast sequence numbers */
2071 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2072 l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2077 if ((msg_tol = msg_link_tolerance(msg)))
2078 link_set_supervision_props(l_ptr, msg_tol);
2080 if (msg_linkprio(msg) &&
2081 (msg_linkprio(msg) != l_ptr->priority)) {
2082 warn("Resetting link <%s>, priority change %u->%u\n",
2083 l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2084 l_ptr->priority = msg_linkprio(msg);
2085 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2088 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2089 l_ptr->stats.recv_states++;
2090 if (link_reset_unknown(l_ptr))
2093 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2094 rec_gap = mod(msg_next_sent(msg) -
2095 mod(l_ptr->next_in_no));
2098 max_pkt_ack = msg_max_pkt(msg);
2099 if (max_pkt_ack > l_ptr->max_pkt) {
2100 l_ptr->max_pkt = max_pkt_ack;
2101 l_ptr->max_pkt_probes = 0;
2105 if (msg_probe(msg)) {
2106 l_ptr->stats.recv_probes++;
2107 if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2108 max_pkt_ack = msg_size(msg);
2112 /* Protocol message before retransmits, reduce loss risk */
2114 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2116 if (rec_gap || (msg_probe(msg))) {
2117 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2118 0, rec_gap, 0, 0, max_pkt_ack);
2120 if (msg_seq_gap(msg)) {
2121 l_ptr->stats.recv_nacks++;
2122 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2133 * tipc_link_tunnel(): Send one message via a link belonging to
2134 * another bearer. Owner node is locked.
2136 static void tipc_link_tunnel(struct link *l_ptr,
2137 struct tipc_msg *tunnel_hdr,
2138 struct tipc_msg *msg,
2141 struct link *tunnel;
2142 struct sk_buff *buf;
2143 u32 length = msg_size(msg);
2145 tunnel = l_ptr->owner->active_links[selector & 1];
2146 if (!tipc_link_is_up(tunnel)) {
2147 warn("Link changeover error, "
2148 "tunnel link no longer available\n");
2151 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2152 buf = tipc_buf_acquire(length + INT_H_SIZE);
2154 warn("Link changeover error, "
2155 "unable to send tunnel msg\n");
2158 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2159 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2160 tipc_link_send_buf(tunnel, buf);
2166 * changeover(): Send whole message queue via the remaining link
2167 * Owner node is locked.
2170 void tipc_link_changeover(struct link *l_ptr)
2172 u32 msgcount = l_ptr->out_queue_size;
2173 struct sk_buff *crs = l_ptr->first_out;
2174 struct link *tunnel = l_ptr->owner->active_links[0];
2175 struct tipc_msg tunnel_hdr;
2181 if (!l_ptr->owner->permit_changeover) {
2182 warn("Link changeover error, "
2183 "peer did not permit changeover\n");
2187 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2188 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2189 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2190 msg_set_msgcnt(&tunnel_hdr, msgcount);
2192 if (!l_ptr->first_out) {
2193 struct sk_buff *buf;
2195 buf = tipc_buf_acquire(INT_H_SIZE);
2197 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2198 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2199 tipc_link_send_buf(tunnel, buf);
2201 warn("Link changeover error, "
2202 "unable to send changeover msg\n");
2207 split_bundles = (l_ptr->owner->active_links[0] !=
2208 l_ptr->owner->active_links[1]);
2211 struct tipc_msg *msg = buf_msg(crs);
2213 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2214 struct tipc_msg *m = msg_get_wrapped(msg);
2215 unchar *pos = (unchar *)m;
2217 msgcount = msg_msgcnt(msg);
2218 while (msgcount--) {
2219 msg_set_seqno(m, msg_seqno(msg));
2220 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2221 msg_link_selector(m));
2222 pos += align(msg_size(m));
2223 m = (struct tipc_msg *)pos;
2226 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2227 msg_link_selector(msg));
2233 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2235 struct sk_buff *iter;
2236 struct tipc_msg tunnel_hdr;
2238 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2239 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2240 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2241 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2242 iter = l_ptr->first_out;
2244 struct sk_buff *outbuf;
2245 struct tipc_msg *msg = buf_msg(iter);
2246 u32 length = msg_size(msg);
2248 if (msg_user(msg) == MSG_BUNDLER)
2249 msg_set_type(msg, CLOSED_MSG);
2250 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2251 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2252 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2253 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2254 if (outbuf == NULL) {
2255 warn("Link changeover error, "
2256 "unable to send duplicate msg\n");
2259 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2260 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2262 tipc_link_send_buf(tunnel, outbuf);
2263 if (!tipc_link_is_up(l_ptr))
2272 * buf_extract - extracts embedded TIPC message from another message
2273 * @skb: encapsulating message buffer
2274 * @from_pos: offset to extract from
2276 * Returns a new message buffer containing an embedded message. The
2277 * encapsulating message itself is left unchanged.
2280 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2282 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2283 u32 size = msg_size(msg);
2286 eb = tipc_buf_acquire(size);
2288 skb_copy_to_linear_data(eb, msg, size);
2293 * link_recv_changeover_msg(): Receive tunneled packet sent
2294 * via other link. Node is locked. Return extracted buffer.
2297 static int link_recv_changeover_msg(struct link **l_ptr,
2298 struct sk_buff **buf)
2300 struct sk_buff *tunnel_buf = *buf;
2301 struct link *dest_link;
2302 struct tipc_msg *msg;
2303 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2304 u32 msg_typ = msg_type(tunnel_msg);
2305 u32 msg_count = msg_msgcnt(tunnel_msg);
2307 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2310 if (dest_link == *l_ptr) {
2311 err("Unexpected changeover message on link <%s>\n",
2316 msg = msg_get_wrapped(tunnel_msg);
2318 if (msg_typ == DUPLICATE_MSG) {
2319 if (less(msg_seqno(msg), mod(dest_link->next_in_no)))
2321 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2323 warn("Link changeover error, duplicate msg dropped\n");
2326 buf_discard(tunnel_buf);
2330 /* First original message ?: */
2332 if (tipc_link_is_up(dest_link)) {
2333 info("Resetting link <%s>, changeover initiated by peer\n",
2335 tipc_link_reset(dest_link);
2336 dest_link->exp_msg_count = msg_count;
2339 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2340 dest_link->exp_msg_count = msg_count;
2345 /* Receive original message */
2347 if (dest_link->exp_msg_count == 0) {
2348 warn("Link switchover error, "
2349 "got too many tunnelled messages\n");
2352 dest_link->exp_msg_count--;
2353 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2356 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2358 buf_discard(tunnel_buf);
2361 warn("Link changeover error, original msg dropped\n");
2366 buf_discard(tunnel_buf);
2371 * Bundler functionality:
2373 void tipc_link_recv_bundle(struct sk_buff *buf)
2375 u32 msgcount = msg_msgcnt(buf_msg(buf));
2376 u32 pos = INT_H_SIZE;
2377 struct sk_buff *obuf;
2379 while (msgcount--) {
2380 obuf = buf_extract(buf, pos);
2382 warn("Link unable to unbundle message(s)\n");
2385 pos += align(msg_size(buf_msg(obuf)));
2386 tipc_net_route_msg(obuf);
2392 * Fragmentation/defragmentation:
2397 * link_send_long_buf: Entry for buffers needing fragmentation.
2398 * The buffer is complete, inclusive total message length.
2399 * Returns user data length.
2401 static int link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2403 struct tipc_msg *inmsg = buf_msg(buf);
2404 struct tipc_msg fragm_hdr;
2405 u32 insize = msg_size(inmsg);
2406 u32 dsz = msg_data_sz(inmsg);
2407 unchar *crs = buf->data;
2409 u32 pack_sz = l_ptr->max_pkt;
2410 u32 fragm_sz = pack_sz - INT_H_SIZE;
2414 if (msg_short(inmsg))
2415 destaddr = l_ptr->addr;
2417 destaddr = msg_destnode(inmsg);
2419 if (msg_routed(inmsg))
2420 msg_set_prevnode(inmsg, tipc_own_addr);
2422 /* Prepare reusable fragment header: */
2424 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2425 INT_H_SIZE, destaddr);
2426 msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2427 msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2428 msg_set_fragm_no(&fragm_hdr, fragm_no);
2429 l_ptr->stats.sent_fragmented++;
2431 /* Chop up message: */
2434 struct sk_buff *fragm;
2436 if (rest <= fragm_sz) {
2438 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2440 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2441 if (fragm == NULL) {
2442 warn("Link unable to fragment message\n");
2446 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2447 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2448 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2450 /* Send queued messages first, if any: */
2452 l_ptr->stats.sent_fragments++;
2453 tipc_link_send_buf(l_ptr, fragm);
2454 if (!tipc_link_is_up(l_ptr))
2456 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2459 msg_set_type(&fragm_hdr, FRAGMENT);
2467 * A pending message being re-assembled must store certain values
2468 * to handle subsequent fragments correctly. The following functions
2469 * help storing these values in unused, available fields in the
2470 * pending message. This makes dynamic memory allocation unecessary.
2473 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2475 msg_set_seqno(buf_msg(buf), seqno);
2478 static u32 get_fragm_size(struct sk_buff *buf)
2480 return msg_ack(buf_msg(buf));
2483 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2485 msg_set_ack(buf_msg(buf), sz);
2488 static u32 get_expected_frags(struct sk_buff *buf)
2490 return msg_bcast_ack(buf_msg(buf));
2493 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2495 msg_set_bcast_ack(buf_msg(buf), exp);
2498 static u32 get_timer_cnt(struct sk_buff *buf)
2500 return msg_reroute_cnt(buf_msg(buf));
2503 static void incr_timer_cnt(struct sk_buff *buf)
2505 msg_incr_reroute_cnt(buf_msg(buf));
2509 * tipc_link_recv_fragment(): Called with node lock on. Returns
2510 * the reassembled buffer if message is complete.
2512 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2513 struct tipc_msg **m)
2515 struct sk_buff *prev = NULL;
2516 struct sk_buff *fbuf = *fb;
2517 struct tipc_msg *fragm = buf_msg(fbuf);
2518 struct sk_buff *pbuf = *pending;
2519 u32 long_msg_seq_no = msg_long_msgno(fragm);
2523 /* Is there an incomplete message waiting for this fragment? */
2525 while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no) ||
2526 (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2531 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2532 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2533 u32 msg_sz = msg_size(imsg);
2534 u32 fragm_sz = msg_data_sz(fragm);
2535 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2536 u32 max = TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2537 if (msg_type(imsg) == TIPC_MCAST_MSG)
2538 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2539 if (msg_size(imsg) > max) {
2543 pbuf = tipc_buf_acquire(msg_size(imsg));
2545 pbuf->next = *pending;
2547 skb_copy_to_linear_data(pbuf, imsg,
2548 msg_data_sz(fragm));
2549 /* Prepare buffer for subsequent fragments. */
2551 set_long_msg_seqno(pbuf, long_msg_seq_no);
2552 set_fragm_size(pbuf, fragm_sz);
2553 set_expected_frags(pbuf, exp_fragm_cnt - 1);
2555 warn("Link unable to reassemble fragmented message\n");
2559 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2560 u32 dsz = msg_data_sz(fragm);
2561 u32 fsz = get_fragm_size(pbuf);
2562 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2563 u32 exp_frags = get_expected_frags(pbuf) - 1;
2564 skb_copy_to_linear_data_offset(pbuf, crs,
2565 msg_data(fragm), dsz);
2568 /* Is message complete? */
2570 if (exp_frags == 0) {
2572 prev->next = pbuf->next;
2574 *pending = pbuf->next;
2575 msg_reset_reroute_cnt(buf_msg(pbuf));
2580 set_expected_frags(pbuf, exp_frags);
2588 * link_check_defragm_bufs - flush stale incoming message fragments
2589 * @l_ptr: pointer to link
2592 static void link_check_defragm_bufs(struct link *l_ptr)
2594 struct sk_buff *prev = NULL;
2595 struct sk_buff *next = NULL;
2596 struct sk_buff *buf = l_ptr->defragm_buf;
2600 if (!link_working_working(l_ptr))
2603 u32 cnt = get_timer_cnt(buf);
2607 incr_timer_cnt(buf);
2611 prev->next = buf->next;
2613 l_ptr->defragm_buf = buf->next;
2622 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2624 l_ptr->tolerance = tolerance;
2625 l_ptr->continuity_interval =
2626 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2627 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2631 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2633 /* Data messages from this node, inclusive FIRST_FRAGM */
2634 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2635 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2636 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2637 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2638 /* Transiting data messages,inclusive FIRST_FRAGM */
2639 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2640 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2641 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2642 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2643 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2644 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2645 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2646 /* FRAGMENT and LAST_FRAGMENT packets */
2647 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2651 * link_find_link - locate link by name
2652 * @name - ptr to link name string
2653 * @node - ptr to area to be filled with ptr to associated node
2655 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2656 * this also prevents link deletion.
2658 * Returns pointer to link (or 0 if invalid link name).
2661 static struct link *link_find_link(const char *name, struct tipc_node **node)
2663 struct link_name link_name_parts;
2664 struct bearer *b_ptr;
2667 if (!link_name_validate(name, &link_name_parts))
2670 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2674 *node = tipc_node_find(link_name_parts.addr_peer);
2678 l_ptr = (*node)->links[b_ptr->identity];
2679 if (!l_ptr || strcmp(l_ptr->name, name))
2685 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2688 struct tipc_link_config *args;
2691 struct tipc_node *node;
2694 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2695 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2697 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2698 new_value = ntohl(args->value);
2700 if (!strcmp(args->name, tipc_bclink_name)) {
2701 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2702 (tipc_bclink_set_queue_limits(new_value) == 0))
2703 return tipc_cfg_reply_none();
2704 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2705 " (cannot change setting on broadcast link)");
2708 read_lock_bh(&tipc_net_lock);
2709 l_ptr = link_find_link(args->name, &node);
2711 read_unlock_bh(&tipc_net_lock);
2712 return tipc_cfg_reply_error_string("link not found");
2715 tipc_node_lock(node);
2718 case TIPC_CMD_SET_LINK_TOL:
2719 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2720 (new_value <= TIPC_MAX_LINK_TOL)) {
2721 link_set_supervision_props(l_ptr, new_value);
2722 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2723 0, 0, new_value, 0, 0);
2727 case TIPC_CMD_SET_LINK_PRI:
2728 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2729 (new_value <= TIPC_MAX_LINK_PRI)) {
2730 l_ptr->priority = new_value;
2731 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2732 0, 0, 0, new_value, 0);
2736 case TIPC_CMD_SET_LINK_WINDOW:
2737 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2738 (new_value <= TIPC_MAX_LINK_WIN)) {
2739 tipc_link_set_queue_limits(l_ptr, new_value);
2744 tipc_node_unlock(node);
2746 read_unlock_bh(&tipc_net_lock);
2748 return tipc_cfg_reply_error_string("cannot change link setting");
2750 return tipc_cfg_reply_none();
2754 * link_reset_statistics - reset link statistics
2755 * @l_ptr: pointer to link
2758 static void link_reset_statistics(struct link *l_ptr)
2760 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2761 l_ptr->stats.sent_info = l_ptr->next_out_no;
2762 l_ptr->stats.recv_info = l_ptr->next_in_no;
2765 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2769 struct tipc_node *node;
2771 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2772 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2774 link_name = (char *)TLV_DATA(req_tlv_area);
2775 if (!strcmp(link_name, tipc_bclink_name)) {
2776 if (tipc_bclink_reset_stats())
2777 return tipc_cfg_reply_error_string("link not found");
2778 return tipc_cfg_reply_none();
2781 read_lock_bh(&tipc_net_lock);
2782 l_ptr = link_find_link(link_name, &node);
2784 read_unlock_bh(&tipc_net_lock);
2785 return tipc_cfg_reply_error_string("link not found");
2788 tipc_node_lock(node);
2789 link_reset_statistics(l_ptr);
2790 tipc_node_unlock(node);
2791 read_unlock_bh(&tipc_net_lock);
2792 return tipc_cfg_reply_none();
2796 * percent - convert count to a percentage of total (rounding up or down)
2799 static u32 percent(u32 count, u32 total)
2801 return (count * 100 + (total / 2)) / total;
2805 * tipc_link_stats - print link statistics
2807 * @buf: print buffer area
2808 * @buf_size: size of print buffer area
2810 * Returns length of print buffer data string (or 0 if error)
2813 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2815 struct print_buf pb;
2817 struct tipc_node *node;
2819 u32 profile_total = 0;
2821 if (!strcmp(name, tipc_bclink_name))
2822 return tipc_bclink_stats(buf, buf_size);
2824 tipc_printbuf_init(&pb, buf, buf_size);
2826 read_lock_bh(&tipc_net_lock);
2827 l_ptr = link_find_link(name, &node);
2829 read_unlock_bh(&tipc_net_lock);
2832 tipc_node_lock(node);
2834 if (tipc_link_is_active(l_ptr))
2836 else if (tipc_link_is_up(l_ptr))
2840 tipc_printf(&pb, "Link <%s>\n"
2841 " %s MTU:%u Priority:%u Tolerance:%u ms"
2842 " Window:%u packets\n",
2843 l_ptr->name, status, l_ptr->max_pkt,
2844 l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
2845 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2846 l_ptr->next_in_no - l_ptr->stats.recv_info,
2847 l_ptr->stats.recv_fragments,
2848 l_ptr->stats.recv_fragmented,
2849 l_ptr->stats.recv_bundles,
2850 l_ptr->stats.recv_bundled);
2851 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2852 l_ptr->next_out_no - l_ptr->stats.sent_info,
2853 l_ptr->stats.sent_fragments,
2854 l_ptr->stats.sent_fragmented,
2855 l_ptr->stats.sent_bundles,
2856 l_ptr->stats.sent_bundled);
2857 profile_total = l_ptr->stats.msg_length_counts;
2860 tipc_printf(&pb, " TX profile sample:%u packets average:%u octets\n"
2861 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2862 "-16354:%u%% -32768:%u%% -66000:%u%%\n",
2863 l_ptr->stats.msg_length_counts,
2864 l_ptr->stats.msg_lengths_total / profile_total,
2865 percent(l_ptr->stats.msg_length_profile[0], profile_total),
2866 percent(l_ptr->stats.msg_length_profile[1], profile_total),
2867 percent(l_ptr->stats.msg_length_profile[2], profile_total),
2868 percent(l_ptr->stats.msg_length_profile[3], profile_total),
2869 percent(l_ptr->stats.msg_length_profile[4], profile_total),
2870 percent(l_ptr->stats.msg_length_profile[5], profile_total),
2871 percent(l_ptr->stats.msg_length_profile[6], profile_total));
2872 tipc_printf(&pb, " RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
2873 l_ptr->stats.recv_states,
2874 l_ptr->stats.recv_probes,
2875 l_ptr->stats.recv_nacks,
2876 l_ptr->stats.deferred_recv,
2877 l_ptr->stats.duplicates);
2878 tipc_printf(&pb, " TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
2879 l_ptr->stats.sent_states,
2880 l_ptr->stats.sent_probes,
2881 l_ptr->stats.sent_nacks,
2882 l_ptr->stats.sent_acks,
2883 l_ptr->stats.retransmitted);
2884 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
2885 l_ptr->stats.bearer_congs,
2886 l_ptr->stats.link_congs,
2887 l_ptr->stats.max_queue_sz,
2888 l_ptr->stats.queue_sz_counts
2889 ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
2892 tipc_node_unlock(node);
2893 read_unlock_bh(&tipc_net_lock);
2894 return tipc_printbuf_validate(&pb);
2897 #define MAX_LINK_STATS_INFO 2000
2899 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2901 struct sk_buff *buf;
2902 struct tlv_desc *rep_tlv;
2905 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2906 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2908 buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
2912 rep_tlv = (struct tlv_desc *)buf->data;
2914 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2915 (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
2918 return tipc_cfg_reply_error_string("link not found");
2921 skb_put(buf, TLV_SPACE(str_len));
2922 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2928 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2929 * @dest: network address of destination node
2930 * @selector: used to select from set of active links
2932 * If no active link can be found, uses default maximum packet size.
2935 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2937 struct tipc_node *n_ptr;
2939 u32 res = MAX_PKT_DEFAULT;
2941 if (dest == tipc_own_addr)
2942 return MAX_MSG_SIZE;
2944 read_lock_bh(&tipc_net_lock);
2945 n_ptr = tipc_node_find(dest);
2947 tipc_node_lock(n_ptr);
2948 l_ptr = n_ptr->active_links[selector & 1];
2950 res = l_ptr->max_pkt;
2951 tipc_node_unlock(n_ptr);
2953 read_unlock_bh(&tipc_net_lock);
2957 static void link_print(struct link *l_ptr, const char *str)
2959 char print_area[256];
2960 struct print_buf pb;
2961 struct print_buf *buf = &pb;
2963 tipc_printbuf_init(buf, print_area, sizeof(print_area));
2965 tipc_printf(buf, str);
2966 tipc_printf(buf, "Link %x<%s>:",
2967 l_ptr->addr, l_ptr->b_ptr->publ.name);
2969 #ifdef CONFIG_TIPC_DEBUG
2970 if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
2973 tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
2974 tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
2975 tipc_printf(buf, "SQUE");
2976 if (l_ptr->first_out) {
2977 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
2978 if (l_ptr->next_out)
2979 tipc_printf(buf, "%u..",
2980 msg_seqno(buf_msg(l_ptr->next_out)));
2981 tipc_printf(buf, "%u]", msg_seqno(buf_msg(l_ptr->last_out)));
2982 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
2983 msg_seqno(buf_msg(l_ptr->first_out)))
2984 != (l_ptr->out_queue_size - 1)) ||
2985 (l_ptr->last_out->next != NULL)) {
2986 tipc_printf(buf, "\nSend queue inconsistency\n");
2987 tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
2988 tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
2989 tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
2992 tipc_printf(buf, "[]");
2993 tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
2994 if (l_ptr->oldest_deferred_in) {
2995 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2996 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
2997 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
2998 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
2999 tipc_printf(buf, ":RQSIZ(%u)",
3000 l_ptr->deferred_inqueue_sz);
3006 if (link_working_unknown(l_ptr))
3007 tipc_printf(buf, ":WU");
3008 else if (link_reset_reset(l_ptr))
3009 tipc_printf(buf, ":RR");
3010 else if (link_reset_unknown(l_ptr))
3011 tipc_printf(buf, ":RU");
3012 else if (link_working_working(l_ptr))
3013 tipc_printf(buf, ":WW");
3014 tipc_printf(buf, "\n");
3016 tipc_printbuf_validate(buf);
3017 info("%s", print_area);