2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2011, 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"
45 * Error message prefixes
47 static const char *link_co_err = "Link changeover error, ";
48 static const char *link_rst_msg = "Resetting link ";
49 static const char *link_unk_evt = "Unknown link event ";
52 * Out-of-range value for link session numbers
54 #define INVALID_SESSION 0x10000
59 #define STARTING_EVT 856384768 /* link processing trigger */
60 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
61 #define TIMEOUT_EVT 560817u /* link timer expired */
64 * The following two 'message types' is really just implementation
65 * data conveniently stored in the message header.
66 * They must not be considered part of the protocol
72 * State value stored in 'exp_msg_count'
74 #define START_CHANGEOVER 100000u
77 * struct tipc_link_name - deconstructed link name
78 * @addr_local: network address of node at this end
79 * @if_local: name of interface at this end
80 * @addr_peer: network address of node at far end
81 * @if_peer: name of interface at far end
83 struct tipc_link_name {
85 char if_local[TIPC_MAX_IF_NAME];
87 char if_peer[TIPC_MAX_IF_NAME];
90 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
92 static void link_recv_proto_msg(struct tipc_link *l_ptr, struct sk_buff *buf);
93 static int link_recv_changeover_msg(struct tipc_link **l_ptr,
94 struct sk_buff **buf);
95 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
96 static int link_send_sections_long(struct tipc_port *sender,
97 struct iovec const *msg_sect,
98 u32 num_sect, unsigned int total_len,
100 static void link_check_defragm_bufs(struct tipc_link *l_ptr);
101 static void link_state_event(struct tipc_link *l_ptr, u32 event);
102 static void link_reset_statistics(struct tipc_link *l_ptr);
103 static void link_print(struct tipc_link *l_ptr, const char *str);
104 static void link_start(struct tipc_link *l_ptr);
105 static int link_send_long_buf(struct tipc_link *l_ptr, struct sk_buff *buf);
108 * Simple link routines
110 static unsigned int align(unsigned int i)
112 return (i + 3) & ~3u;
115 static void link_init_max_pkt(struct tipc_link *l_ptr)
119 max_pkt = (l_ptr->b_ptr->mtu & ~3);
120 if (max_pkt > MAX_MSG_SIZE)
121 max_pkt = MAX_MSG_SIZE;
123 l_ptr->max_pkt_target = max_pkt;
124 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
125 l_ptr->max_pkt = l_ptr->max_pkt_target;
127 l_ptr->max_pkt = MAX_PKT_DEFAULT;
129 l_ptr->max_pkt_probes = 0;
132 static u32 link_next_sent(struct tipc_link *l_ptr)
135 return buf_seqno(l_ptr->next_out);
136 return mod(l_ptr->next_out_no);
139 static u32 link_last_sent(struct tipc_link *l_ptr)
141 return mod(link_next_sent(l_ptr) - 1);
145 * Simple non-static link routines (i.e. referenced outside this file)
147 int tipc_link_is_up(struct tipc_link *l_ptr)
151 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
154 int tipc_link_is_active(struct tipc_link *l_ptr)
156 return (l_ptr->owner->active_links[0] == l_ptr) ||
157 (l_ptr->owner->active_links[1] == l_ptr);
161 * link_name_validate - validate & (optionally) deconstruct tipc_link name
162 * @name: ptr to link name string
163 * @name_parts: ptr to area for link name components (or NULL if not needed)
165 * Returns 1 if link name is valid, otherwise 0.
167 static int link_name_validate(const char *name,
168 struct tipc_link_name *name_parts)
170 char name_copy[TIPC_MAX_LINK_NAME];
176 u32 z_local, c_local, n_local;
177 u32 z_peer, c_peer, n_peer;
181 /* copy link name & ensure length is OK */
182 name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
183 /* need above in case non-Posix strncpy() doesn't pad with nulls */
184 strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
185 if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
188 /* ensure all component parts of link name are present */
189 addr_local = name_copy;
190 if_local = strchr(addr_local, ':');
191 if (if_local == NULL)
194 addr_peer = strchr(if_local, '-');
195 if (addr_peer == NULL)
198 if_local_len = addr_peer - if_local;
199 if_peer = strchr(addr_peer, ':');
203 if_peer_len = strlen(if_peer) + 1;
205 /* validate component parts of link name */
206 if ((sscanf(addr_local, "%u.%u.%u%c",
207 &z_local, &c_local, &n_local, &dummy) != 3) ||
208 (sscanf(addr_peer, "%u.%u.%u%c",
209 &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
210 (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
211 (z_peer > 255) || (c_peer > 4095) || (n_peer > 4095) ||
212 (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
213 (if_peer_len <= 1) || (if_peer_len > TIPC_MAX_IF_NAME))
216 /* return link name components, if necessary */
218 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
219 strcpy(name_parts->if_local, if_local);
220 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
221 strcpy(name_parts->if_peer, if_peer);
227 * link_timeout - handle expiration of link timer
228 * @l_ptr: pointer to link
230 * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
231 * with tipc_link_delete(). (There is no risk that the node will be deleted by
232 * another thread because tipc_link_delete() always cancels the link timer before
233 * tipc_node_delete() is called.)
235 static void link_timeout(struct tipc_link *l_ptr)
237 tipc_node_lock(l_ptr->owner);
239 /* update counters used in statistical profiling of send traffic */
240 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
241 l_ptr->stats.queue_sz_counts++;
243 if (l_ptr->first_out) {
244 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
245 u32 length = msg_size(msg);
247 if ((msg_user(msg) == MSG_FRAGMENTER) &&
248 (msg_type(msg) == FIRST_FRAGMENT)) {
249 length = msg_size(msg_get_wrapped(msg));
252 l_ptr->stats.msg_lengths_total += length;
253 l_ptr->stats.msg_length_counts++;
255 l_ptr->stats.msg_length_profile[0]++;
256 else if (length <= 256)
257 l_ptr->stats.msg_length_profile[1]++;
258 else if (length <= 1024)
259 l_ptr->stats.msg_length_profile[2]++;
260 else if (length <= 4096)
261 l_ptr->stats.msg_length_profile[3]++;
262 else if (length <= 16384)
263 l_ptr->stats.msg_length_profile[4]++;
264 else if (length <= 32768)
265 l_ptr->stats.msg_length_profile[5]++;
267 l_ptr->stats.msg_length_profile[6]++;
271 /* do all other link processing performed on a periodic basis */
272 link_check_defragm_bufs(l_ptr);
274 link_state_event(l_ptr, TIMEOUT_EVT);
277 tipc_link_push_queue(l_ptr);
279 tipc_node_unlock(l_ptr->owner);
282 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
284 k_start_timer(&l_ptr->timer, time);
288 * tipc_link_create - create a new link
289 * @n_ptr: pointer to associated node
290 * @b_ptr: pointer to associated bearer
291 * @media_addr: media address to use when sending messages over link
293 * Returns pointer to link.
295 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
296 struct tipc_bearer *b_ptr,
297 const struct tipc_media_addr *media_addr)
299 struct tipc_link *l_ptr;
300 struct tipc_msg *msg;
302 char addr_string[16];
303 u32 peer = n_ptr->addr;
305 if (n_ptr->link_cnt >= 2) {
306 tipc_addr_string_fill(addr_string, n_ptr->addr);
307 pr_err("Attempt to establish third link to %s\n", addr_string);
311 if (n_ptr->links[b_ptr->identity]) {
312 tipc_addr_string_fill(addr_string, n_ptr->addr);
313 pr_err("Attempt to establish second link on <%s> to %s\n",
314 b_ptr->name, addr_string);
318 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
320 pr_warn("Link creation failed, no memory\n");
325 if_name = strchr(b_ptr->name, ':') + 1;
326 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
327 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
328 tipc_node(tipc_own_addr),
330 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
331 /* note: peer i/f name is updated by reset/activate message */
332 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
333 l_ptr->owner = n_ptr;
334 l_ptr->checkpoint = 1;
335 l_ptr->peer_session = INVALID_SESSION;
336 l_ptr->b_ptr = b_ptr;
337 link_set_supervision_props(l_ptr, b_ptr->tolerance);
338 l_ptr->state = RESET_UNKNOWN;
340 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
342 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
343 msg_set_size(msg, sizeof(l_ptr->proto_msg));
344 msg_set_session(msg, (tipc_random & 0xffff));
345 msg_set_bearer_id(msg, b_ptr->identity);
346 strcpy((char *)msg_data(msg), if_name);
348 l_ptr->priority = b_ptr->priority;
349 tipc_link_set_queue_limits(l_ptr, b_ptr->window);
351 link_init_max_pkt(l_ptr);
353 l_ptr->next_out_no = 1;
354 INIT_LIST_HEAD(&l_ptr->waiting_ports);
356 link_reset_statistics(l_ptr);
358 tipc_node_attach_link(n_ptr, l_ptr);
360 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
361 list_add_tail(&l_ptr->link_list, &b_ptr->links);
362 tipc_k_signal((Handler)link_start, (unsigned long)l_ptr);
368 * tipc_link_delete - delete a link
369 * @l_ptr: pointer to link
371 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
372 * This routine must not grab the node lock until after link timer cancellation
373 * to avoid a potential deadlock situation.
375 void tipc_link_delete(struct tipc_link *l_ptr)
378 pr_err("Attempt to delete non-existent link\n");
382 k_cancel_timer(&l_ptr->timer);
384 tipc_node_lock(l_ptr->owner);
385 tipc_link_reset(l_ptr);
386 tipc_node_detach_link(l_ptr->owner, l_ptr);
387 tipc_link_stop(l_ptr);
388 list_del_init(&l_ptr->link_list);
389 tipc_node_unlock(l_ptr->owner);
390 k_term_timer(&l_ptr->timer);
394 static void link_start(struct tipc_link *l_ptr)
396 tipc_node_lock(l_ptr->owner);
397 link_state_event(l_ptr, STARTING_EVT);
398 tipc_node_unlock(l_ptr->owner);
402 * link_schedule_port - schedule port for deferred sending
403 * @l_ptr: pointer to link
404 * @origport: reference to sending port
405 * @sz: amount of data to be sent
407 * Schedules port for renewed sending of messages after link congestion
410 static int link_schedule_port(struct tipc_link *l_ptr, u32 origport, u32 sz)
412 struct tipc_port *p_ptr;
414 spin_lock_bh(&tipc_port_list_lock);
415 p_ptr = tipc_port_lock(origport);
419 if (!list_empty(&p_ptr->wait_list))
421 p_ptr->congested = 1;
422 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
423 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
424 l_ptr->stats.link_congs++;
426 tipc_port_unlock(p_ptr);
428 spin_unlock_bh(&tipc_port_list_lock);
432 void tipc_link_wakeup_ports(struct tipc_link *l_ptr, int all)
434 struct tipc_port *p_ptr;
435 struct tipc_port *temp_p_ptr;
436 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
442 if (!spin_trylock_bh(&tipc_port_list_lock))
444 if (link_congested(l_ptr))
446 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
450 list_del_init(&p_ptr->wait_list);
451 spin_lock_bh(p_ptr->lock);
452 p_ptr->congested = 0;
453 p_ptr->wakeup(p_ptr);
454 win -= p_ptr->waiting_pkts;
455 spin_unlock_bh(p_ptr->lock);
459 spin_unlock_bh(&tipc_port_list_lock);
463 * link_release_outqueue - purge link's outbound message queue
464 * @l_ptr: pointer to link
466 static void link_release_outqueue(struct tipc_link *l_ptr)
468 struct sk_buff *buf = l_ptr->first_out;
469 struct sk_buff *next;
476 l_ptr->first_out = NULL;
477 l_ptr->out_queue_size = 0;
481 * tipc_link_reset_fragments - purge link's inbound message fragments queue
482 * @l_ptr: pointer to link
484 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
486 struct sk_buff *buf = l_ptr->defragm_buf;
487 struct sk_buff *next;
494 l_ptr->defragm_buf = NULL;
498 * tipc_link_stop - purge all inbound and outbound messages associated with link
499 * @l_ptr: pointer to link
501 void tipc_link_stop(struct tipc_link *l_ptr)
504 struct sk_buff *next;
506 buf = l_ptr->oldest_deferred_in;
513 buf = l_ptr->first_out;
520 tipc_link_reset_fragments(l_ptr);
522 kfree_skb(l_ptr->proto_msg_queue);
523 l_ptr->proto_msg_queue = NULL;
526 void tipc_link_reset(struct tipc_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_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: */
556 link_release_outqueue(l_ptr);
557 kfree_skb(l_ptr->proto_msg_queue);
558 l_ptr->proto_msg_queue = NULL;
559 buf = l_ptr->oldest_deferred_in;
561 struct sk_buff *next = buf->next;
565 if (!list_empty(&l_ptr->waiting_ports))
566 tipc_link_wakeup_ports(l_ptr, 1);
568 l_ptr->retransm_queue_head = 0;
569 l_ptr->retransm_queue_size = 0;
570 l_ptr->last_out = NULL;
571 l_ptr->first_out = NULL;
572 l_ptr->next_out = NULL;
573 l_ptr->unacked_window = 0;
574 l_ptr->checkpoint = 1;
575 l_ptr->next_out_no = 1;
576 l_ptr->deferred_inqueue_sz = 0;
577 l_ptr->oldest_deferred_in = NULL;
578 l_ptr->newest_deferred_in = NULL;
579 l_ptr->fsm_msg_cnt = 0;
580 l_ptr->stale_count = 0;
581 link_reset_statistics(l_ptr);
585 static void link_activate(struct tipc_link *l_ptr)
587 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
588 tipc_node_link_up(l_ptr->owner, l_ptr);
589 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
593 * link_state_event - link finite state machine
594 * @l_ptr: pointer to link
595 * @event: state machine event to process
597 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
599 struct tipc_link *other;
600 u32 cont_intv = l_ptr->continuity_interval;
602 if (!l_ptr->started && (event != STARTING_EVT))
603 return; /* Not yet. */
605 if (link_blocked(l_ptr)) {
606 if (event == TIMEOUT_EVT)
607 link_set_timer(l_ptr, cont_intv);
608 return; /* Changeover going on */
611 switch (l_ptr->state) {
612 case WORKING_WORKING:
614 case TRAFFIC_MSG_EVT:
618 if (l_ptr->next_in_no != l_ptr->checkpoint) {
619 l_ptr->checkpoint = l_ptr->next_in_no;
620 if (tipc_bclink_acks_missing(l_ptr->owner)) {
621 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
623 l_ptr->fsm_msg_cnt++;
624 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
625 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
627 l_ptr->fsm_msg_cnt++;
629 link_set_timer(l_ptr, cont_intv);
632 l_ptr->state = WORKING_UNKNOWN;
633 l_ptr->fsm_msg_cnt = 0;
634 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
635 l_ptr->fsm_msg_cnt++;
636 link_set_timer(l_ptr, cont_intv / 4);
639 pr_info("%s<%s>, requested by peer\n", link_rst_msg,
641 tipc_link_reset(l_ptr);
642 l_ptr->state = RESET_RESET;
643 l_ptr->fsm_msg_cnt = 0;
644 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
645 l_ptr->fsm_msg_cnt++;
646 link_set_timer(l_ptr, cont_intv);
649 pr_err("%s%u in WW state\n", link_unk_evt, event);
652 case WORKING_UNKNOWN:
654 case TRAFFIC_MSG_EVT:
656 l_ptr->state = WORKING_WORKING;
657 l_ptr->fsm_msg_cnt = 0;
658 link_set_timer(l_ptr, cont_intv);
661 pr_info("%s<%s>, requested by peer while probing\n",
662 link_rst_msg, l_ptr->name);
663 tipc_link_reset(l_ptr);
664 l_ptr->state = RESET_RESET;
665 l_ptr->fsm_msg_cnt = 0;
666 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
667 l_ptr->fsm_msg_cnt++;
668 link_set_timer(l_ptr, cont_intv);
671 if (l_ptr->next_in_no != l_ptr->checkpoint) {
672 l_ptr->state = WORKING_WORKING;
673 l_ptr->fsm_msg_cnt = 0;
674 l_ptr->checkpoint = l_ptr->next_in_no;
675 if (tipc_bclink_acks_missing(l_ptr->owner)) {
676 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
678 l_ptr->fsm_msg_cnt++;
680 link_set_timer(l_ptr, cont_intv);
681 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
682 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
684 l_ptr->fsm_msg_cnt++;
685 link_set_timer(l_ptr, cont_intv / 4);
686 } else { /* Link has failed */
687 pr_warn("%s<%s>, peer not responding\n",
688 link_rst_msg, l_ptr->name);
689 tipc_link_reset(l_ptr);
690 l_ptr->state = RESET_UNKNOWN;
691 l_ptr->fsm_msg_cnt = 0;
692 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
694 l_ptr->fsm_msg_cnt++;
695 link_set_timer(l_ptr, cont_intv);
699 pr_err("%s%u in WU state\n", link_unk_evt, event);
704 case TRAFFIC_MSG_EVT:
707 other = l_ptr->owner->active_links[0];
708 if (other && link_working_unknown(other))
710 l_ptr->state = WORKING_WORKING;
711 l_ptr->fsm_msg_cnt = 0;
712 link_activate(l_ptr);
713 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
714 l_ptr->fsm_msg_cnt++;
715 link_set_timer(l_ptr, cont_intv);
718 l_ptr->state = RESET_RESET;
719 l_ptr->fsm_msg_cnt = 0;
720 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
721 l_ptr->fsm_msg_cnt++;
722 link_set_timer(l_ptr, cont_intv);
728 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
729 l_ptr->fsm_msg_cnt++;
730 link_set_timer(l_ptr, cont_intv);
733 pr_err("%s%u in RU state\n", link_unk_evt, event);
738 case TRAFFIC_MSG_EVT:
740 other = l_ptr->owner->active_links[0];
741 if (other && link_working_unknown(other))
743 l_ptr->state = WORKING_WORKING;
744 l_ptr->fsm_msg_cnt = 0;
745 link_activate(l_ptr);
746 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
747 l_ptr->fsm_msg_cnt++;
748 link_set_timer(l_ptr, cont_intv);
753 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
754 l_ptr->fsm_msg_cnt++;
755 link_set_timer(l_ptr, cont_intv);
758 pr_err("%s%u in RR state\n", link_unk_evt, event);
762 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
767 * link_bundle_buf(): Append contents of a buffer to
768 * the tail of an existing one.
770 static int link_bundle_buf(struct tipc_link *l_ptr,
771 struct sk_buff *bundler,
774 struct tipc_msg *bundler_msg = buf_msg(bundler);
775 struct tipc_msg *msg = buf_msg(buf);
776 u32 size = msg_size(msg);
777 u32 bundle_size = msg_size(bundler_msg);
778 u32 to_pos = align(bundle_size);
779 u32 pad = to_pos - bundle_size;
781 if (msg_user(bundler_msg) != MSG_BUNDLER)
783 if (msg_type(bundler_msg) != OPEN_MSG)
785 if (skb_tailroom(bundler) < (pad + size))
787 if (l_ptr->max_pkt < (to_pos + size))
790 skb_put(bundler, pad + size);
791 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
792 msg_set_size(bundler_msg, to_pos + size);
793 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
795 l_ptr->stats.sent_bundled++;
799 static void link_add_to_outqueue(struct tipc_link *l_ptr,
801 struct tipc_msg *msg)
803 u32 ack = mod(l_ptr->next_in_no - 1);
804 u32 seqno = mod(l_ptr->next_out_no++);
806 msg_set_word(msg, 2, ((ack << 16) | seqno));
807 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
809 if (l_ptr->first_out) {
810 l_ptr->last_out->next = buf;
811 l_ptr->last_out = buf;
813 l_ptr->first_out = l_ptr->last_out = buf;
815 l_ptr->out_queue_size++;
816 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
817 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
820 static void link_add_chain_to_outqueue(struct tipc_link *l_ptr,
821 struct sk_buff *buf_chain,
825 struct tipc_msg *msg;
827 if (!l_ptr->next_out)
828 l_ptr->next_out = buf_chain;
831 buf_chain = buf_chain->next;
834 msg_set_long_msgno(msg, long_msgno);
835 link_add_to_outqueue(l_ptr, buf, msg);
840 * tipc_link_send_buf() is the 'full path' for messages, called from
841 * inside TIPC when the 'fast path' in tipc_send_buf
842 * has failed, and from link_send()
844 int tipc_link_send_buf(struct tipc_link *l_ptr, struct sk_buff *buf)
846 struct tipc_msg *msg = buf_msg(buf);
847 u32 size = msg_size(msg);
848 u32 dsz = msg_data_sz(msg);
849 u32 queue_size = l_ptr->out_queue_size;
850 u32 imp = tipc_msg_tot_importance(msg);
851 u32 queue_limit = l_ptr->queue_limit[imp];
852 u32 max_packet = l_ptr->max_pkt;
854 /* Match msg importance against queue limits: */
855 if (unlikely(queue_size >= queue_limit)) {
856 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
857 link_schedule_port(l_ptr, msg_origport(msg), size);
862 if (imp > CONN_MANAGER) {
863 pr_warn("%s<%s>, send queue full", link_rst_msg,
865 tipc_link_reset(l_ptr);
870 /* Fragmentation needed ? */
871 if (size > max_packet)
872 return link_send_long_buf(l_ptr, buf);
874 /* Packet can be queued or sent. */
875 if (likely(!tipc_bearer_blocked(l_ptr->b_ptr) &&
876 !link_congested(l_ptr))) {
877 link_add_to_outqueue(l_ptr, buf, msg);
879 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
880 l_ptr->unacked_window = 0;
883 /* Congestion: can message be bundled ? */
884 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
885 (msg_user(msg) != MSG_FRAGMENTER)) {
887 /* Try adding message to an existing bundle */
888 if (l_ptr->next_out &&
889 link_bundle_buf(l_ptr, l_ptr->last_out, buf))
892 /* Try creating a new bundle */
893 if (size <= max_packet * 2 / 3) {
894 struct sk_buff *bundler = tipc_buf_acquire(max_packet);
895 struct tipc_msg bundler_hdr;
898 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
899 INT_H_SIZE, l_ptr->addr);
900 skb_copy_to_linear_data(bundler, &bundler_hdr,
902 skb_trim(bundler, INT_H_SIZE);
903 link_bundle_buf(l_ptr, bundler, buf);
906 l_ptr->stats.sent_bundles++;
910 if (!l_ptr->next_out)
911 l_ptr->next_out = buf;
912 link_add_to_outqueue(l_ptr, buf, msg);
917 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
918 * not been selected yet, and the the owner node is not locked
919 * Called by TIPC internal users, e.g. the name distributor
921 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
923 struct tipc_link *l_ptr;
924 struct tipc_node *n_ptr;
925 int res = -ELINKCONG;
927 read_lock_bh(&tipc_net_lock);
928 n_ptr = tipc_node_find(dest);
930 tipc_node_lock(n_ptr);
931 l_ptr = n_ptr->active_links[selector & 1];
933 res = tipc_link_send_buf(l_ptr, buf);
936 tipc_node_unlock(n_ptr);
940 read_unlock_bh(&tipc_net_lock);
945 * tipc_link_send_names - send name table entries to new neighbor
947 * Send routine for bulk delivery of name table messages when contact
948 * with a new neighbor occurs. No link congestion checking is performed
949 * because name table messages *must* be delivered. The messages must be
950 * small enough not to require fragmentation.
951 * Called without any locks held.
953 void tipc_link_send_names(struct list_head *message_list, u32 dest)
955 struct tipc_node *n_ptr;
956 struct tipc_link *l_ptr;
958 struct sk_buff *temp_buf;
960 if (list_empty(message_list))
963 read_lock_bh(&tipc_net_lock);
964 n_ptr = tipc_node_find(dest);
966 tipc_node_lock(n_ptr);
967 l_ptr = n_ptr->active_links[0];
969 /* convert circular list to linear list */
970 ((struct sk_buff *)message_list->prev)->next = NULL;
971 link_add_chain_to_outqueue(l_ptr,
972 (struct sk_buff *)message_list->next, 0);
973 tipc_link_push_queue(l_ptr);
974 INIT_LIST_HEAD(message_list);
976 tipc_node_unlock(n_ptr);
978 read_unlock_bh(&tipc_net_lock);
980 /* discard the messages if they couldn't be sent */
981 list_for_each_safe(buf, temp_buf, ((struct sk_buff *)message_list)) {
982 list_del((struct list_head *)buf);
988 * link_send_buf_fast: Entry for data messages where the
989 * destination link is known and the header is complete,
990 * inclusive total message length. Very time critical.
991 * Link is locked. Returns user data length.
993 static int link_send_buf_fast(struct tipc_link *l_ptr, struct sk_buff *buf,
996 struct tipc_msg *msg = buf_msg(buf);
997 int res = msg_data_sz(msg);
999 if (likely(!link_congested(l_ptr))) {
1000 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
1001 if (likely(!tipc_bearer_blocked(l_ptr->b_ptr))) {
1002 link_add_to_outqueue(l_ptr, buf, msg);
1003 tipc_bearer_send(l_ptr->b_ptr, buf,
1004 &l_ptr->media_addr);
1005 l_ptr->unacked_window = 0;
1009 *used_max_pkt = l_ptr->max_pkt;
1011 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
1015 * tipc_send_buf_fast: Entry for data messages where the
1016 * destination node is known and the header is complete,
1017 * inclusive total message length.
1018 * Returns user data length.
1020 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1022 struct tipc_link *l_ptr;
1023 struct tipc_node *n_ptr;
1025 u32 selector = msg_origport(buf_msg(buf)) & 1;
1028 read_lock_bh(&tipc_net_lock);
1029 n_ptr = tipc_node_find(destnode);
1030 if (likely(n_ptr)) {
1031 tipc_node_lock(n_ptr);
1032 l_ptr = n_ptr->active_links[selector];
1033 if (likely(l_ptr)) {
1034 res = link_send_buf_fast(l_ptr, buf, &dummy);
1035 tipc_node_unlock(n_ptr);
1036 read_unlock_bh(&tipc_net_lock);
1039 tipc_node_unlock(n_ptr);
1041 read_unlock_bh(&tipc_net_lock);
1042 res = msg_data_sz(buf_msg(buf));
1043 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1049 * tipc_link_send_sections_fast: Entry for messages where the
1050 * destination processor is known and the header is complete,
1051 * except for total message length.
1052 * Returns user data length or errno.
1054 int tipc_link_send_sections_fast(struct tipc_port *sender,
1055 struct iovec const *msg_sect,
1057 unsigned int total_len,
1060 struct tipc_msg *hdr = &sender->phdr;
1061 struct tipc_link *l_ptr;
1062 struct sk_buff *buf;
1063 struct tipc_node *node;
1065 u32 selector = msg_origport(hdr) & 1;
1069 * Try building message using port's max_pkt hint.
1070 * (Must not hold any locks while building message.)
1072 res = tipc_msg_build(hdr, msg_sect, num_sect, total_len,
1073 sender->max_pkt, !sender->user_port, &buf);
1075 read_lock_bh(&tipc_net_lock);
1076 node = tipc_node_find(destaddr);
1078 tipc_node_lock(node);
1079 l_ptr = node->active_links[selector];
1080 if (likely(l_ptr)) {
1082 res = link_send_buf_fast(l_ptr, buf,
1085 tipc_node_unlock(node);
1086 read_unlock_bh(&tipc_net_lock);
1090 /* Exit if build request was invalid */
1091 if (unlikely(res < 0))
1094 /* Exit if link (or bearer) is congested */
1095 if (link_congested(l_ptr) ||
1096 tipc_bearer_blocked(l_ptr->b_ptr)) {
1097 res = link_schedule_port(l_ptr,
1103 * Message size exceeds max_pkt hint; update hint,
1104 * then re-try fast path or fragment the message
1106 sender->max_pkt = l_ptr->max_pkt;
1107 tipc_node_unlock(node);
1108 read_unlock_bh(&tipc_net_lock);
1111 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1114 return link_send_sections_long(sender, msg_sect,
1115 num_sect, total_len,
1118 tipc_node_unlock(node);
1120 read_unlock_bh(&tipc_net_lock);
1122 /* Couldn't find a link to the destination node */
1124 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1126 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1127 total_len, TIPC_ERR_NO_NODE);
1132 * link_send_sections_long(): Entry for long messages where the
1133 * destination node is known and the header is complete,
1134 * inclusive total message length.
1135 * Link and bearer congestion status have been checked to be ok,
1136 * and are ignored if they change.
1138 * Note that fragments do not use the full link MTU so that they won't have
1139 * to undergo refragmentation if link changeover causes them to be sent
1140 * over another link with an additional tunnel header added as prefix.
1141 * (Refragmentation will still occur if the other link has a smaller MTU.)
1143 * Returns user data length or errno.
1145 static int link_send_sections_long(struct tipc_port *sender,
1146 struct iovec const *msg_sect,
1148 unsigned int total_len,
1151 struct tipc_link *l_ptr;
1152 struct tipc_node *node;
1153 struct tipc_msg *hdr = &sender->phdr;
1154 u32 dsz = total_len;
1155 u32 max_pkt, fragm_sz, rest;
1156 struct tipc_msg fragm_hdr;
1157 struct sk_buff *buf, *buf_chain, *prev;
1158 u32 fragm_crs, fragm_rest, hsz, sect_rest;
1159 const unchar *sect_crs;
1165 max_pkt = sender->max_pkt - INT_H_SIZE;
1166 /* leave room for tunnel header in case of link changeover */
1167 fragm_sz = max_pkt - INT_H_SIZE;
1168 /* leave room for fragmentation header in each fragment */
1176 /* Prepare reusable fragment header */
1177 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1178 INT_H_SIZE, msg_destnode(hdr));
1179 msg_set_size(&fragm_hdr, max_pkt);
1180 msg_set_fragm_no(&fragm_hdr, 1);
1182 /* Prepare header of first fragment */
1183 buf_chain = buf = tipc_buf_acquire(max_pkt);
1187 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1188 hsz = msg_hdr_sz(hdr);
1189 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1191 /* Chop up message */
1192 fragm_crs = INT_H_SIZE + hsz;
1193 fragm_rest = fragm_sz - hsz;
1195 do { /* For all sections */
1199 sect_rest = msg_sect[++curr_sect].iov_len;
1200 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1203 if (sect_rest < fragm_rest)
1208 if (likely(!sender->user_port)) {
1209 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1211 for (; buf_chain; buf_chain = buf) {
1212 buf = buf_chain->next;
1213 kfree_skb(buf_chain);
1218 skb_copy_to_linear_data_offset(buf, fragm_crs,
1226 if (!fragm_rest && rest) {
1228 /* Initiate new fragment: */
1229 if (rest <= fragm_sz) {
1231 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1233 msg_set_type(&fragm_hdr, FRAGMENT);
1235 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1236 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1238 buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1244 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1245 fragm_crs = INT_H_SIZE;
1246 fragm_rest = fragm_sz;
1251 * Now we have a buffer chain. Select a link and check
1252 * that packet size is still OK
1254 node = tipc_node_find(destaddr);
1256 tipc_node_lock(node);
1257 l_ptr = node->active_links[sender->ref & 1];
1259 tipc_node_unlock(node);
1262 if (l_ptr->max_pkt < max_pkt) {
1263 sender->max_pkt = l_ptr->max_pkt;
1264 tipc_node_unlock(node);
1265 for (; buf_chain; buf_chain = buf) {
1266 buf = buf_chain->next;
1267 kfree_skb(buf_chain);
1273 for (; buf_chain; buf_chain = buf) {
1274 buf = buf_chain->next;
1275 kfree_skb(buf_chain);
1277 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1278 total_len, TIPC_ERR_NO_NODE);
1281 /* Append chain of fragments to send queue & send them */
1282 l_ptr->long_msg_seq_no++;
1283 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1284 l_ptr->stats.sent_fragments += fragm_no;
1285 l_ptr->stats.sent_fragmented++;
1286 tipc_link_push_queue(l_ptr);
1287 tipc_node_unlock(node);
1292 * tipc_link_push_packet: Push one unsent packet to the media
1294 u32 tipc_link_push_packet(struct tipc_link *l_ptr)
1296 struct sk_buff *buf = l_ptr->first_out;
1297 u32 r_q_size = l_ptr->retransm_queue_size;
1298 u32 r_q_head = l_ptr->retransm_queue_head;
1300 /* Step to position where retransmission failed, if any, */
1301 /* consider that buffers may have been released in meantime */
1302 if (r_q_size && buf) {
1303 u32 last = lesser(mod(r_q_head + r_q_size),
1304 link_last_sent(l_ptr));
1305 u32 first = buf_seqno(buf);
1307 while (buf && less(first, r_q_head)) {
1308 first = mod(first + 1);
1311 l_ptr->retransm_queue_head = r_q_head = first;
1312 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1315 /* Continue retransmission now, if there is anything: */
1316 if (r_q_size && buf) {
1317 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1318 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1319 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1320 l_ptr->retransm_queue_head = mod(++r_q_head);
1321 l_ptr->retransm_queue_size = --r_q_size;
1322 l_ptr->stats.retransmitted++;
1326 /* Send deferred protocol message, if any: */
1327 buf = l_ptr->proto_msg_queue;
1329 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1330 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1331 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1332 l_ptr->unacked_window = 0;
1334 l_ptr->proto_msg_queue = NULL;
1338 /* Send one deferred data message, if send window not full: */
1339 buf = l_ptr->next_out;
1341 struct tipc_msg *msg = buf_msg(buf);
1342 u32 next = msg_seqno(msg);
1343 u32 first = buf_seqno(l_ptr->first_out);
1345 if (mod(next - first) < l_ptr->queue_limit[0]) {
1346 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1347 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1348 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1349 if (msg_user(msg) == MSG_BUNDLER)
1350 msg_set_type(msg, CLOSED_MSG);
1351 l_ptr->next_out = buf->next;
1359 * push_queue(): push out the unsent messages of a link where
1360 * congestion has abated. Node is locked
1362 void tipc_link_push_queue(struct tipc_link *l_ptr)
1366 if (tipc_bearer_blocked(l_ptr->b_ptr))
1370 res = tipc_link_push_packet(l_ptr);
1374 static void link_reset_all(unsigned long addr)
1376 struct tipc_node *n_ptr;
1377 char addr_string[16];
1380 read_lock_bh(&tipc_net_lock);
1381 n_ptr = tipc_node_find((u32)addr);
1383 read_unlock_bh(&tipc_net_lock);
1384 return; /* node no longer exists */
1387 tipc_node_lock(n_ptr);
1389 pr_warn("Resetting all links to %s\n",
1390 tipc_addr_string_fill(addr_string, n_ptr->addr));
1392 for (i = 0; i < MAX_BEARERS; i++) {
1393 if (n_ptr->links[i]) {
1394 link_print(n_ptr->links[i], "Resetting link\n");
1395 tipc_link_reset(n_ptr->links[i]);
1399 tipc_node_unlock(n_ptr);
1400 read_unlock_bh(&tipc_net_lock);
1403 static void link_retransmit_failure(struct tipc_link *l_ptr,
1404 struct sk_buff *buf)
1406 struct tipc_msg *msg = buf_msg(buf);
1408 pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
1411 /* Handle failure on standard link */
1412 link_print(l_ptr, "Resetting link\n");
1413 tipc_link_reset(l_ptr);
1416 /* Handle failure on broadcast link */
1417 struct tipc_node *n_ptr;
1418 char addr_string[16];
1420 pr_info("Msg seq number: %u, ", msg_seqno(msg));
1421 pr_cont("Outstanding acks: %lu\n",
1422 (unsigned long) TIPC_SKB_CB(buf)->handle);
1424 n_ptr = tipc_bclink_retransmit_to();
1425 tipc_node_lock(n_ptr);
1427 tipc_addr_string_fill(addr_string, n_ptr->addr);
1428 pr_info("Broadcast link info for %s\n", addr_string);
1429 pr_info("Reception permitted: %d, Acked: %u\n",
1430 n_ptr->bclink.recv_permitted,
1431 n_ptr->bclink.acked);
1432 pr_info("Last in: %u, Oos state: %u, Last sent: %u\n",
1433 n_ptr->bclink.last_in,
1434 n_ptr->bclink.oos_state,
1435 n_ptr->bclink.last_sent);
1437 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1439 tipc_node_unlock(n_ptr);
1441 l_ptr->stale_count = 0;
1445 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
1448 struct tipc_msg *msg;
1455 if (tipc_bearer_blocked(l_ptr->b_ptr)) {
1456 if (l_ptr->retransm_queue_size == 0) {
1457 l_ptr->retransm_queue_head = msg_seqno(msg);
1458 l_ptr->retransm_queue_size = retransmits;
1460 pr_err("Unexpected retransmit on link %s (qsize=%d)\n",
1461 l_ptr->name, l_ptr->retransm_queue_size);
1465 /* Detect repeated retransmit failures on unblocked bearer */
1466 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1467 if (++l_ptr->stale_count > 100) {
1468 link_retransmit_failure(l_ptr, buf);
1472 l_ptr->last_retransmitted = msg_seqno(msg);
1473 l_ptr->stale_count = 1;
1477 while (retransmits && (buf != l_ptr->next_out) && buf) {
1479 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1480 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1481 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1484 l_ptr->stats.retransmitted++;
1487 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1491 * link_insert_deferred_queue - insert deferred messages back into receive chain
1493 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1494 struct sk_buff *buf)
1498 if (l_ptr->oldest_deferred_in == NULL)
1501 seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1502 if (seq_no == mod(l_ptr->next_in_no)) {
1503 l_ptr->newest_deferred_in->next = buf;
1504 buf = l_ptr->oldest_deferred_in;
1505 l_ptr->oldest_deferred_in = NULL;
1506 l_ptr->deferred_inqueue_sz = 0;
1512 * link_recv_buf_validate - validate basic format of received message
1514 * This routine ensures a TIPC message has an acceptable header, and at least
1515 * as much data as the header indicates it should. The routine also ensures
1516 * that the entire message header is stored in the main fragment of the message
1517 * buffer, to simplify future access to message header fields.
1519 * Note: Having extra info present in the message header or data areas is OK.
1520 * TIPC will ignore the excess, under the assumption that it is optional info
1521 * introduced by a later release of the protocol.
1523 static int link_recv_buf_validate(struct sk_buff *buf)
1525 static u32 min_data_hdr_size[8] = {
1526 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1527 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1530 struct tipc_msg *msg;
1536 if (unlikely(buf->len < MIN_H_SIZE))
1539 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1543 if (unlikely(msg_version(msg) != TIPC_VERSION))
1546 size = msg_size(msg);
1547 hdr_size = msg_hdr_sz(msg);
1548 min_hdr_size = msg_isdata(msg) ?
1549 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1551 if (unlikely((hdr_size < min_hdr_size) ||
1552 (size < hdr_size) ||
1553 (buf->len < size) ||
1554 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1557 return pskb_may_pull(buf, hdr_size);
1561 * tipc_recv_msg - process TIPC messages arriving from off-node
1562 * @head: pointer to message buffer chain
1563 * @tb_ptr: pointer to bearer message arrived on
1565 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1566 * structure (i.e. cannot be NULL), but bearer can be inactive.
1568 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *b_ptr)
1570 read_lock_bh(&tipc_net_lock);
1572 struct tipc_node *n_ptr;
1573 struct tipc_link *l_ptr;
1574 struct sk_buff *crs;
1575 struct sk_buff *buf = head;
1576 struct tipc_msg *msg;
1584 /* Ensure bearer is still enabled */
1585 if (unlikely(!b_ptr->active))
1588 /* Ensure message is well-formed */
1589 if (unlikely(!link_recv_buf_validate(buf)))
1592 /* Ensure message data is a single contiguous unit */
1593 if (unlikely(skb_linearize(buf)))
1596 /* Handle arrival of a non-unicast link message */
1599 if (unlikely(msg_non_seq(msg))) {
1600 if (msg_user(msg) == LINK_CONFIG)
1601 tipc_disc_recv_msg(buf, b_ptr);
1603 tipc_bclink_recv_pkt(buf);
1607 /* Discard unicast link messages destined for another node */
1608 if (unlikely(!msg_short(msg) &&
1609 (msg_destnode(msg) != tipc_own_addr)))
1612 /* Locate neighboring node that sent message */
1613 n_ptr = tipc_node_find(msg_prevnode(msg));
1614 if (unlikely(!n_ptr))
1616 tipc_node_lock(n_ptr);
1618 /* Locate unicast link endpoint that should handle message */
1619 l_ptr = n_ptr->links[b_ptr->identity];
1620 if (unlikely(!l_ptr)) {
1621 tipc_node_unlock(n_ptr);
1625 /* Verify that communication with node is currently allowed */
1626 if ((n_ptr->block_setup & WAIT_PEER_DOWN) &&
1627 msg_user(msg) == LINK_PROTOCOL &&
1628 (msg_type(msg) == RESET_MSG ||
1629 msg_type(msg) == ACTIVATE_MSG) &&
1630 !msg_redundant_link(msg))
1631 n_ptr->block_setup &= ~WAIT_PEER_DOWN;
1633 if (n_ptr->block_setup) {
1634 tipc_node_unlock(n_ptr);
1638 /* Validate message sequence number info */
1639 seq_no = msg_seqno(msg);
1640 ackd = msg_ack(msg);
1642 /* Release acked messages */
1643 if (n_ptr->bclink.recv_permitted)
1644 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1646 crs = l_ptr->first_out;
1647 while ((crs != l_ptr->next_out) &&
1648 less_eq(buf_seqno(crs), ackd)) {
1649 struct sk_buff *next = crs->next;
1656 l_ptr->first_out = crs;
1657 l_ptr->out_queue_size -= released;
1660 /* Try sending any messages link endpoint has pending */
1661 if (unlikely(l_ptr->next_out))
1662 tipc_link_push_queue(l_ptr);
1663 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1664 tipc_link_wakeup_ports(l_ptr, 0);
1665 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1666 l_ptr->stats.sent_acks++;
1667 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1670 /* Now (finally!) process the incoming message */
1672 if (likely(link_working_working(l_ptr))) {
1673 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1674 l_ptr->next_in_no++;
1675 if (unlikely(l_ptr->oldest_deferred_in))
1676 head = link_insert_deferred_queue(l_ptr,
1679 if (likely(msg_isdata(msg))) {
1680 tipc_node_unlock(n_ptr);
1681 tipc_port_recv_msg(buf);
1684 switch (msg_user(msg)) {
1687 l_ptr->stats.recv_bundles++;
1688 l_ptr->stats.recv_bundled +=
1690 tipc_node_unlock(n_ptr);
1691 tipc_link_recv_bundle(buf);
1693 case NAME_DISTRIBUTOR:
1694 tipc_node_unlock(n_ptr);
1695 tipc_named_recv(buf);
1698 tipc_node_unlock(n_ptr);
1699 tipc_port_recv_proto_msg(buf);
1701 case MSG_FRAGMENTER:
1702 l_ptr->stats.recv_fragments++;
1703 ret = tipc_link_recv_fragment(
1704 &l_ptr->defragm_buf,
1707 l_ptr->stats.recv_fragmented++;
1711 l_ptr->next_in_no--;
1713 case CHANGEOVER_PROTOCOL:
1714 type = msg_type(msg);
1715 if (link_recv_changeover_msg(&l_ptr,
1718 seq_no = msg_seqno(msg);
1719 if (type == ORIGINAL_MSG)
1721 goto protocol_check;
1729 tipc_node_unlock(n_ptr);
1730 tipc_net_route_msg(buf);
1733 link_handle_out_of_seq_msg(l_ptr, buf);
1734 head = link_insert_deferred_queue(l_ptr, head);
1735 tipc_node_unlock(n_ptr);
1739 if (msg_user(msg) == LINK_PROTOCOL) {
1740 link_recv_proto_msg(l_ptr, buf);
1741 head = link_insert_deferred_queue(l_ptr, head);
1742 tipc_node_unlock(n_ptr);
1745 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1747 if (link_working_working(l_ptr)) {
1748 /* Re-insert in front of queue */
1751 tipc_node_unlock(n_ptr);
1754 tipc_node_unlock(n_ptr);
1758 read_unlock_bh(&tipc_net_lock);
1762 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1764 * Returns increase in queue length (i.e. 0 or 1)
1766 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1767 struct sk_buff *buf)
1769 struct sk_buff *queue_buf;
1770 struct sk_buff **prev;
1771 u32 seq_no = buf_seqno(buf);
1776 if (*head == NULL) {
1777 *head = *tail = buf;
1782 if (less(buf_seqno(*tail), seq_no)) {
1783 (*tail)->next = buf;
1788 /* Locate insertion point in queue, then insert; discard if duplicate */
1792 u32 curr_seqno = buf_seqno(queue_buf);
1794 if (seq_no == curr_seqno) {
1799 if (less(seq_no, curr_seqno))
1802 prev = &queue_buf->next;
1803 queue_buf = queue_buf->next;
1806 buf->next = queue_buf;
1812 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1814 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1815 struct sk_buff *buf)
1817 u32 seq_no = buf_seqno(buf);
1819 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1820 link_recv_proto_msg(l_ptr, buf);
1824 /* Record OOS packet arrival (force mismatch on next timeout) */
1825 l_ptr->checkpoint--;
1828 * Discard packet if a duplicate; otherwise add it to deferred queue
1829 * and notify peer of gap as per protocol specification
1831 if (less(seq_no, mod(l_ptr->next_in_no))) {
1832 l_ptr->stats.duplicates++;
1837 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1838 &l_ptr->newest_deferred_in, buf)) {
1839 l_ptr->deferred_inqueue_sz++;
1840 l_ptr->stats.deferred_recv++;
1841 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1842 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1844 l_ptr->stats.duplicates++;
1848 * Send protocol message to the other endpoint.
1850 void tipc_link_send_proto_msg(struct tipc_link *l_ptr, u32 msg_typ,
1851 int probe_msg, u32 gap, u32 tolerance,
1852 u32 priority, u32 ack_mtu)
1854 struct sk_buff *buf = NULL;
1855 struct tipc_msg *msg = l_ptr->pmsg;
1856 u32 msg_size = sizeof(l_ptr->proto_msg);
1859 /* Discard any previous message that was deferred due to congestion */
1860 if (l_ptr->proto_msg_queue) {
1861 kfree_skb(l_ptr->proto_msg_queue);
1862 l_ptr->proto_msg_queue = NULL;
1865 if (link_blocked(l_ptr))
1868 /* Abort non-RESET send if communication with node is prohibited */
1869 if ((l_ptr->owner->block_setup) && (msg_typ != RESET_MSG))
1872 /* Create protocol message with "out-of-sequence" sequence number */
1873 msg_set_type(msg, msg_typ);
1874 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1875 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1876 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1878 if (msg_typ == STATE_MSG) {
1879 u32 next_sent = mod(l_ptr->next_out_no);
1881 if (!tipc_link_is_up(l_ptr))
1883 if (l_ptr->next_out)
1884 next_sent = buf_seqno(l_ptr->next_out);
1885 msg_set_next_sent(msg, next_sent);
1886 if (l_ptr->oldest_deferred_in) {
1887 u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1888 gap = mod(rec - mod(l_ptr->next_in_no));
1890 msg_set_seq_gap(msg, gap);
1892 l_ptr->stats.sent_nacks++;
1893 msg_set_link_tolerance(msg, tolerance);
1894 msg_set_linkprio(msg, priority);
1895 msg_set_max_pkt(msg, ack_mtu);
1896 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1897 msg_set_probe(msg, probe_msg != 0);
1899 u32 mtu = l_ptr->max_pkt;
1901 if ((mtu < l_ptr->max_pkt_target) &&
1902 link_working_working(l_ptr) &&
1903 l_ptr->fsm_msg_cnt) {
1904 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1905 if (l_ptr->max_pkt_probes == 10) {
1906 l_ptr->max_pkt_target = (msg_size - 4);
1907 l_ptr->max_pkt_probes = 0;
1908 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1910 l_ptr->max_pkt_probes++;
1913 l_ptr->stats.sent_probes++;
1915 l_ptr->stats.sent_states++;
1916 } else { /* RESET_MSG or ACTIVATE_MSG */
1917 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1918 msg_set_seq_gap(msg, 0);
1919 msg_set_next_sent(msg, 1);
1920 msg_set_probe(msg, 0);
1921 msg_set_link_tolerance(msg, l_ptr->tolerance);
1922 msg_set_linkprio(msg, l_ptr->priority);
1923 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1926 r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1927 msg_set_redundant_link(msg, r_flag);
1928 msg_set_linkprio(msg, l_ptr->priority);
1929 msg_set_size(msg, msg_size);
1931 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1933 buf = tipc_buf_acquire(msg_size);
1937 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1939 /* Defer message if bearer is already blocked */
1940 if (tipc_bearer_blocked(l_ptr->b_ptr)) {
1941 l_ptr->proto_msg_queue = buf;
1945 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1946 l_ptr->unacked_window = 0;
1951 * Receive protocol message :
1952 * Note that network plane id propagates through the network, and may
1953 * change at any time. The node with lowest address rules
1955 static void link_recv_proto_msg(struct tipc_link *l_ptr, struct sk_buff *buf)
1961 struct tipc_msg *msg = buf_msg(buf);
1963 if (link_blocked(l_ptr))
1966 /* record unnumbered packet arrival (force mismatch on next timeout) */
1967 l_ptr->checkpoint--;
1969 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
1970 if (tipc_own_addr > msg_prevnode(msg))
1971 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
1973 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
1975 switch (msg_type(msg)) {
1978 if (!link_working_unknown(l_ptr) &&
1979 (l_ptr->peer_session != INVALID_SESSION)) {
1980 if (less_eq(msg_session(msg), l_ptr->peer_session))
1981 break; /* duplicate or old reset: ignore */
1984 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1985 link_working_unknown(l_ptr))) {
1987 * peer has lost contact -- don't allow peer's links
1988 * to reactivate before we recognize loss & clean up
1990 l_ptr->owner->block_setup = WAIT_NODE_DOWN;
1993 link_state_event(l_ptr, RESET_MSG);
1997 /* Update link settings according other endpoint's values */
1998 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2000 msg_tol = msg_link_tolerance(msg);
2001 if (msg_tol > l_ptr->tolerance)
2002 link_set_supervision_props(l_ptr, msg_tol);
2004 if (msg_linkprio(msg) > l_ptr->priority)
2005 l_ptr->priority = msg_linkprio(msg);
2007 max_pkt_info = msg_max_pkt(msg);
2009 if (max_pkt_info < l_ptr->max_pkt_target)
2010 l_ptr->max_pkt_target = max_pkt_info;
2011 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2012 l_ptr->max_pkt = l_ptr->max_pkt_target;
2014 l_ptr->max_pkt = l_ptr->max_pkt_target;
2017 /* Synchronize broadcast link info, if not done previously */
2018 if (!tipc_node_is_up(l_ptr->owner)) {
2019 l_ptr->owner->bclink.last_sent =
2020 l_ptr->owner->bclink.last_in =
2021 msg_last_bcast(msg);
2022 l_ptr->owner->bclink.oos_state = 0;
2025 l_ptr->peer_session = msg_session(msg);
2026 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2028 if (msg_type(msg) == ACTIVATE_MSG)
2029 link_state_event(l_ptr, ACTIVATE_MSG);
2033 msg_tol = msg_link_tolerance(msg);
2035 link_set_supervision_props(l_ptr, msg_tol);
2037 if (msg_linkprio(msg) &&
2038 (msg_linkprio(msg) != l_ptr->priority)) {
2039 pr_warn("%s<%s>, priority change %u->%u\n",
2040 link_rst_msg, l_ptr->name, l_ptr->priority,
2042 l_ptr->priority = msg_linkprio(msg);
2043 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2046 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2047 l_ptr->stats.recv_states++;
2048 if (link_reset_unknown(l_ptr))
2051 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2052 rec_gap = mod(msg_next_sent(msg) -
2053 mod(l_ptr->next_in_no));
2056 max_pkt_ack = msg_max_pkt(msg);
2057 if (max_pkt_ack > l_ptr->max_pkt) {
2058 l_ptr->max_pkt = max_pkt_ack;
2059 l_ptr->max_pkt_probes = 0;
2063 if (msg_probe(msg)) {
2064 l_ptr->stats.recv_probes++;
2065 if (msg_size(msg) > sizeof(l_ptr->proto_msg))
2066 max_pkt_ack = msg_size(msg);
2069 /* Protocol message before retransmits, reduce loss risk */
2070 if (l_ptr->owner->bclink.recv_permitted)
2071 tipc_bclink_update_link_state(l_ptr->owner,
2072 msg_last_bcast(msg));
2074 if (rec_gap || (msg_probe(msg))) {
2075 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2076 0, rec_gap, 0, 0, max_pkt_ack);
2078 if (msg_seq_gap(msg)) {
2079 l_ptr->stats.recv_nacks++;
2080 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2091 * tipc_link_tunnel(): Send one message via a link belonging to
2092 * another bearer. Owner node is locked.
2094 static void tipc_link_tunnel(struct tipc_link *l_ptr,
2095 struct tipc_msg *tunnel_hdr,
2096 struct tipc_msg *msg,
2099 struct tipc_link *tunnel;
2100 struct sk_buff *buf;
2101 u32 length = msg_size(msg);
2103 tunnel = l_ptr->owner->active_links[selector & 1];
2104 if (!tipc_link_is_up(tunnel)) {
2105 pr_warn("%stunnel link no longer available\n", link_co_err);
2108 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2109 buf = tipc_buf_acquire(length + INT_H_SIZE);
2111 pr_warn("%sunable to send tunnel msg\n", link_co_err);
2114 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2115 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2116 tipc_link_send_buf(tunnel, buf);
2122 * changeover(): Send whole message queue via the remaining link
2123 * Owner node is locked.
2125 void tipc_link_changeover(struct tipc_link *l_ptr)
2127 u32 msgcount = l_ptr->out_queue_size;
2128 struct sk_buff *crs = l_ptr->first_out;
2129 struct tipc_link *tunnel = l_ptr->owner->active_links[0];
2130 struct tipc_msg tunnel_hdr;
2136 if (!l_ptr->owner->permit_changeover) {
2137 pr_warn("%speer did not permit changeover\n", link_co_err);
2141 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2142 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2143 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2144 msg_set_msgcnt(&tunnel_hdr, msgcount);
2146 if (!l_ptr->first_out) {
2147 struct sk_buff *buf;
2149 buf = tipc_buf_acquire(INT_H_SIZE);
2151 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2152 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2153 tipc_link_send_buf(tunnel, buf);
2155 pr_warn("%sunable to send changeover msg\n",
2161 split_bundles = (l_ptr->owner->active_links[0] !=
2162 l_ptr->owner->active_links[1]);
2165 struct tipc_msg *msg = buf_msg(crs);
2167 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2168 struct tipc_msg *m = msg_get_wrapped(msg);
2169 unchar *pos = (unchar *)m;
2171 msgcount = msg_msgcnt(msg);
2172 while (msgcount--) {
2173 msg_set_seqno(m, msg_seqno(msg));
2174 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2175 msg_link_selector(m));
2176 pos += align(msg_size(m));
2177 m = (struct tipc_msg *)pos;
2180 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2181 msg_link_selector(msg));
2187 void tipc_link_send_duplicate(struct tipc_link *l_ptr, struct tipc_link *tunnel)
2189 struct sk_buff *iter;
2190 struct tipc_msg tunnel_hdr;
2192 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2193 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2194 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2195 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2196 iter = l_ptr->first_out;
2198 struct sk_buff *outbuf;
2199 struct tipc_msg *msg = buf_msg(iter);
2200 u32 length = msg_size(msg);
2202 if (msg_user(msg) == MSG_BUNDLER)
2203 msg_set_type(msg, CLOSED_MSG);
2204 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2205 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2206 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2207 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2208 if (outbuf == NULL) {
2209 pr_warn("%sunable to send duplicate msg\n",
2213 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2214 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2216 tipc_link_send_buf(tunnel, outbuf);
2217 if (!tipc_link_is_up(l_ptr))
2224 * buf_extract - extracts embedded TIPC message from another message
2225 * @skb: encapsulating message buffer
2226 * @from_pos: offset to extract from
2228 * Returns a new message buffer containing an embedded message. The
2229 * encapsulating message itself is left unchanged.
2231 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2233 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2234 u32 size = msg_size(msg);
2237 eb = tipc_buf_acquire(size);
2239 skb_copy_to_linear_data(eb, msg, size);
2244 * link_recv_changeover_msg(): Receive tunneled packet sent
2245 * via other link. Node is locked. Return extracted buffer.
2247 static int link_recv_changeover_msg(struct tipc_link **l_ptr,
2248 struct sk_buff **buf)
2250 struct sk_buff *tunnel_buf = *buf;
2251 struct tipc_link *dest_link;
2252 struct tipc_msg *msg;
2253 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2254 u32 msg_typ = msg_type(tunnel_msg);
2255 u32 msg_count = msg_msgcnt(tunnel_msg);
2257 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2260 if (dest_link == *l_ptr) {
2261 pr_err("Unexpected changeover message on link <%s>\n",
2266 msg = msg_get_wrapped(tunnel_msg);
2268 if (msg_typ == DUPLICATE_MSG) {
2269 if (less(msg_seqno(msg), mod(dest_link->next_in_no)))
2271 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2273 pr_warn("%sduplicate msg dropped\n", link_co_err);
2276 kfree_skb(tunnel_buf);
2280 /* First original message ?: */
2281 if (tipc_link_is_up(dest_link)) {
2282 pr_info("%s<%s>, changeover initiated by peer\n", link_rst_msg,
2284 tipc_link_reset(dest_link);
2285 dest_link->exp_msg_count = msg_count;
2288 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2289 dest_link->exp_msg_count = msg_count;
2294 /* Receive original message */
2295 if (dest_link->exp_msg_count == 0) {
2296 pr_warn("%sgot too many tunnelled messages\n", link_co_err);
2299 dest_link->exp_msg_count--;
2300 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2303 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2305 kfree_skb(tunnel_buf);
2308 pr_warn("%soriginal msg dropped\n", link_co_err);
2313 kfree_skb(tunnel_buf);
2318 * Bundler functionality:
2320 void tipc_link_recv_bundle(struct sk_buff *buf)
2322 u32 msgcount = msg_msgcnt(buf_msg(buf));
2323 u32 pos = INT_H_SIZE;
2324 struct sk_buff *obuf;
2326 while (msgcount--) {
2327 obuf = buf_extract(buf, pos);
2329 pr_warn("Link unable to unbundle message(s)\n");
2332 pos += align(msg_size(buf_msg(obuf)));
2333 tipc_net_route_msg(obuf);
2339 * Fragmentation/defragmentation:
2343 * link_send_long_buf: Entry for buffers needing fragmentation.
2344 * The buffer is complete, inclusive total message length.
2345 * Returns user data length.
2347 static int link_send_long_buf(struct tipc_link *l_ptr, struct sk_buff *buf)
2349 struct sk_buff *buf_chain = NULL;
2350 struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2351 struct tipc_msg *inmsg = buf_msg(buf);
2352 struct tipc_msg fragm_hdr;
2353 u32 insize = msg_size(inmsg);
2354 u32 dsz = msg_data_sz(inmsg);
2355 unchar *crs = buf->data;
2357 u32 pack_sz = l_ptr->max_pkt;
2358 u32 fragm_sz = pack_sz - INT_H_SIZE;
2362 if (msg_short(inmsg))
2363 destaddr = l_ptr->addr;
2365 destaddr = msg_destnode(inmsg);
2367 /* Prepare reusable fragment header: */
2368 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2369 INT_H_SIZE, destaddr);
2371 /* Chop up message: */
2373 struct sk_buff *fragm;
2375 if (rest <= fragm_sz) {
2377 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2379 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2380 if (fragm == NULL) {
2384 buf_chain = buf_chain->next;
2389 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2391 msg_set_fragm_no(&fragm_hdr, fragm_no);
2392 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2393 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2395 buf_chain_tail->next = fragm;
2396 buf_chain_tail = fragm;
2400 msg_set_type(&fragm_hdr, FRAGMENT);
2404 /* Append chain of fragments to send queue & send them */
2405 l_ptr->long_msg_seq_no++;
2406 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2407 l_ptr->stats.sent_fragments += fragm_no;
2408 l_ptr->stats.sent_fragmented++;
2409 tipc_link_push_queue(l_ptr);
2415 * A pending message being re-assembled must store certain values
2416 * to handle subsequent fragments correctly. The following functions
2417 * help storing these values in unused, available fields in the
2418 * pending message. This makes dynamic memory allocation unnecessary.
2420 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2422 msg_set_seqno(buf_msg(buf), seqno);
2425 static u32 get_fragm_size(struct sk_buff *buf)
2427 return msg_ack(buf_msg(buf));
2430 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2432 msg_set_ack(buf_msg(buf), sz);
2435 static u32 get_expected_frags(struct sk_buff *buf)
2437 return msg_bcast_ack(buf_msg(buf));
2440 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2442 msg_set_bcast_ack(buf_msg(buf), exp);
2445 static u32 get_timer_cnt(struct sk_buff *buf)
2447 return msg_reroute_cnt(buf_msg(buf));
2450 static void incr_timer_cnt(struct sk_buff *buf)
2452 msg_incr_reroute_cnt(buf_msg(buf));
2456 * tipc_link_recv_fragment(): Called with node lock on. Returns
2457 * the reassembled buffer if message is complete.
2459 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2460 struct tipc_msg **m)
2462 struct sk_buff *prev = NULL;
2463 struct sk_buff *fbuf = *fb;
2464 struct tipc_msg *fragm = buf_msg(fbuf);
2465 struct sk_buff *pbuf = *pending;
2466 u32 long_msg_seq_no = msg_long_msgno(fragm);
2470 /* Is there an incomplete message waiting for this fragment? */
2471 while (pbuf && ((buf_seqno(pbuf) != long_msg_seq_no) ||
2472 (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2477 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2478 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2479 u32 msg_sz = msg_size(imsg);
2480 u32 fragm_sz = msg_data_sz(fragm);
2481 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2482 u32 max = TIPC_MAX_USER_MSG_SIZE + NAMED_H_SIZE;
2483 if (msg_type(imsg) == TIPC_MCAST_MSG)
2484 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2485 if (msg_size(imsg) > max) {
2489 pbuf = tipc_buf_acquire(msg_size(imsg));
2491 pbuf->next = *pending;
2493 skb_copy_to_linear_data(pbuf, imsg,
2494 msg_data_sz(fragm));
2495 /* Prepare buffer for subsequent fragments. */
2496 set_long_msg_seqno(pbuf, long_msg_seq_no);
2497 set_fragm_size(pbuf, fragm_sz);
2498 set_expected_frags(pbuf, exp_fragm_cnt - 1);
2500 pr_debug("Link unable to reassemble fragmented message\n");
2506 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2507 u32 dsz = msg_data_sz(fragm);
2508 u32 fsz = get_fragm_size(pbuf);
2509 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2510 u32 exp_frags = get_expected_frags(pbuf) - 1;
2511 skb_copy_to_linear_data_offset(pbuf, crs,
2512 msg_data(fragm), dsz);
2515 /* Is message complete? */
2516 if (exp_frags == 0) {
2518 prev->next = pbuf->next;
2520 *pending = pbuf->next;
2521 msg_reset_reroute_cnt(buf_msg(pbuf));
2526 set_expected_frags(pbuf, exp_frags);
2534 * link_check_defragm_bufs - flush stale incoming message fragments
2535 * @l_ptr: pointer to link
2537 static void link_check_defragm_bufs(struct tipc_link *l_ptr)
2539 struct sk_buff *prev = NULL;
2540 struct sk_buff *next = NULL;
2541 struct sk_buff *buf = l_ptr->defragm_buf;
2545 if (!link_working_working(l_ptr))
2548 u32 cnt = get_timer_cnt(buf);
2552 incr_timer_cnt(buf);
2556 prev->next = buf->next;
2558 l_ptr->defragm_buf = buf->next;
2565 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
2567 if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2570 l_ptr->tolerance = tolerance;
2571 l_ptr->continuity_interval =
2572 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2573 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2576 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
2578 /* Data messages from this node, inclusive FIRST_FRAGM */
2579 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2580 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2581 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2582 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2583 /* Transiting data messages,inclusive FIRST_FRAGM */
2584 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2585 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2586 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2587 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2588 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2589 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2590 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2591 /* FRAGMENT and LAST_FRAGMENT packets */
2592 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2596 * link_find_link - locate link by name
2597 * @name: ptr to link name string
2598 * @node: ptr to area to be filled with ptr to associated node
2600 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2601 * this also prevents link deletion.
2603 * Returns pointer to link (or 0 if invalid link name).
2605 static struct tipc_link *link_find_link(const char *name,
2606 struct tipc_node **node)
2608 struct tipc_link_name link_name_parts;
2609 struct tipc_bearer *b_ptr;
2610 struct tipc_link *l_ptr;
2612 if (!link_name_validate(name, &link_name_parts))
2615 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2619 *node = tipc_node_find(link_name_parts.addr_peer);
2623 l_ptr = (*node)->links[b_ptr->identity];
2624 if (!l_ptr || strcmp(l_ptr->name, name))
2631 * link_value_is_valid -- validate proposed link tolerance/priority/window
2633 * @cmd: value type (TIPC_CMD_SET_LINK_*)
2634 * @new_value: the new value
2636 * Returns 1 if value is within range, 0 if not.
2638 static int link_value_is_valid(u16 cmd, u32 new_value)
2641 case TIPC_CMD_SET_LINK_TOL:
2642 return (new_value >= TIPC_MIN_LINK_TOL) &&
2643 (new_value <= TIPC_MAX_LINK_TOL);
2644 case TIPC_CMD_SET_LINK_PRI:
2645 return (new_value <= TIPC_MAX_LINK_PRI);
2646 case TIPC_CMD_SET_LINK_WINDOW:
2647 return (new_value >= TIPC_MIN_LINK_WIN) &&
2648 (new_value <= TIPC_MAX_LINK_WIN);
2654 * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2655 * @name: ptr to link, bearer, or media name
2656 * @new_value: new value of link, bearer, or media setting
2657 * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2659 * Caller must hold 'tipc_net_lock' to ensure link/bearer/media is not deleted.
2661 * Returns 0 if value updated and negative value on error.
2663 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2665 struct tipc_node *node;
2666 struct tipc_link *l_ptr;
2667 struct tipc_bearer *b_ptr;
2668 struct tipc_media *m_ptr;
2670 l_ptr = link_find_link(name, &node);
2673 * acquire node lock for tipc_link_send_proto_msg().
2674 * see "TIPC locking policy" in net.c.
2676 tipc_node_lock(node);
2678 case TIPC_CMD_SET_LINK_TOL:
2679 link_set_supervision_props(l_ptr, new_value);
2680 tipc_link_send_proto_msg(l_ptr,
2681 STATE_MSG, 0, 0, new_value, 0, 0);
2683 case TIPC_CMD_SET_LINK_PRI:
2684 l_ptr->priority = new_value;
2685 tipc_link_send_proto_msg(l_ptr,
2686 STATE_MSG, 0, 0, 0, new_value, 0);
2688 case TIPC_CMD_SET_LINK_WINDOW:
2689 tipc_link_set_queue_limits(l_ptr, new_value);
2692 tipc_node_unlock(node);
2696 b_ptr = tipc_bearer_find(name);
2699 case TIPC_CMD_SET_LINK_TOL:
2700 b_ptr->tolerance = new_value;
2702 case TIPC_CMD_SET_LINK_PRI:
2703 b_ptr->priority = new_value;
2705 case TIPC_CMD_SET_LINK_WINDOW:
2706 b_ptr->window = new_value;
2712 m_ptr = tipc_media_find(name);
2716 case TIPC_CMD_SET_LINK_TOL:
2717 m_ptr->tolerance = new_value;
2719 case TIPC_CMD_SET_LINK_PRI:
2720 m_ptr->priority = new_value;
2722 case TIPC_CMD_SET_LINK_WINDOW:
2723 m_ptr->window = new_value;
2729 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2732 struct tipc_link_config *args;
2736 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2737 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2739 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2740 new_value = ntohl(args->value);
2742 if (!link_value_is_valid(cmd, new_value))
2743 return tipc_cfg_reply_error_string(
2744 "cannot change, value invalid");
2746 if (!strcmp(args->name, tipc_bclink_name)) {
2747 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2748 (tipc_bclink_set_queue_limits(new_value) == 0))
2749 return tipc_cfg_reply_none();
2750 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2751 " (cannot change setting on broadcast link)");
2754 read_lock_bh(&tipc_net_lock);
2755 res = link_cmd_set_value(args->name, new_value, cmd);
2756 read_unlock_bh(&tipc_net_lock);
2758 return tipc_cfg_reply_error_string("cannot change link setting");
2760 return tipc_cfg_reply_none();
2764 * link_reset_statistics - reset link statistics
2765 * @l_ptr: pointer to link
2767 static void link_reset_statistics(struct tipc_link *l_ptr)
2769 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2770 l_ptr->stats.sent_info = l_ptr->next_out_no;
2771 l_ptr->stats.recv_info = l_ptr->next_in_no;
2774 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2777 struct tipc_link *l_ptr;
2778 struct tipc_node *node;
2780 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2781 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2783 link_name = (char *)TLV_DATA(req_tlv_area);
2784 if (!strcmp(link_name, tipc_bclink_name)) {
2785 if (tipc_bclink_reset_stats())
2786 return tipc_cfg_reply_error_string("link not found");
2787 return tipc_cfg_reply_none();
2790 read_lock_bh(&tipc_net_lock);
2791 l_ptr = link_find_link(link_name, &node);
2793 read_unlock_bh(&tipc_net_lock);
2794 return tipc_cfg_reply_error_string("link not found");
2797 tipc_node_lock(node);
2798 link_reset_statistics(l_ptr);
2799 tipc_node_unlock(node);
2800 read_unlock_bh(&tipc_net_lock);
2801 return tipc_cfg_reply_none();
2805 * percent - convert count to a percentage of total (rounding up or down)
2807 static u32 percent(u32 count, u32 total)
2809 return (count * 100 + (total / 2)) / total;
2813 * tipc_link_stats - print link statistics
2815 * @buf: print buffer area
2816 * @buf_size: size of print buffer area
2818 * Returns length of print buffer data string (or 0 if error)
2820 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2822 struct tipc_link *l;
2823 struct tipc_stats *s;
2824 struct tipc_node *node;
2826 u32 profile_total = 0;
2829 if (!strcmp(name, tipc_bclink_name))
2830 return tipc_bclink_stats(buf, buf_size);
2832 read_lock_bh(&tipc_net_lock);
2833 l = link_find_link(name, &node);
2835 read_unlock_bh(&tipc_net_lock);
2838 tipc_node_lock(node);
2841 if (tipc_link_is_active(l))
2843 else if (tipc_link_is_up(l))
2848 ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2849 " %s MTU:%u Priority:%u Tolerance:%u ms"
2850 " Window:%u packets\n",
2851 l->name, status, l->max_pkt, l->priority,
2852 l->tolerance, l->queue_limit[0]);
2854 ret += tipc_snprintf(buf + ret, buf_size - ret,
2855 " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2856 l->next_in_no - s->recv_info, s->recv_fragments,
2857 s->recv_fragmented, s->recv_bundles,
2860 ret += tipc_snprintf(buf + ret, buf_size - ret,
2861 " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2862 l->next_out_no - s->sent_info, s->sent_fragments,
2863 s->sent_fragmented, s->sent_bundles,
2866 profile_total = s->msg_length_counts;
2870 ret += tipc_snprintf(buf + ret, buf_size - ret,
2871 " TX profile sample:%u packets average:%u octets\n"
2872 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2873 "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2874 s->msg_length_counts,
2875 s->msg_lengths_total / profile_total,
2876 percent(s->msg_length_profile[0], profile_total),
2877 percent(s->msg_length_profile[1], profile_total),
2878 percent(s->msg_length_profile[2], profile_total),
2879 percent(s->msg_length_profile[3], profile_total),
2880 percent(s->msg_length_profile[4], profile_total),
2881 percent(s->msg_length_profile[5], profile_total),
2882 percent(s->msg_length_profile[6], profile_total));
2884 ret += tipc_snprintf(buf + ret, buf_size - ret,
2885 " RX states:%u probes:%u naks:%u defs:%u"
2886 " dups:%u\n", s->recv_states, s->recv_probes,
2887 s->recv_nacks, s->deferred_recv, s->duplicates);
2889 ret += tipc_snprintf(buf + ret, buf_size - ret,
2890 " TX states:%u probes:%u naks:%u acks:%u"
2891 " dups:%u\n", s->sent_states, s->sent_probes,
2892 s->sent_nacks, s->sent_acks, s->retransmitted);
2894 ret += tipc_snprintf(buf + ret, buf_size - ret,
2895 " Congestion link:%u Send queue"
2896 " max:%u avg:%u\n", s->link_congs,
2897 s->max_queue_sz, s->queue_sz_counts ?
2898 (s->accu_queue_sz / s->queue_sz_counts) : 0);
2900 tipc_node_unlock(node);
2901 read_unlock_bh(&tipc_net_lock);
2905 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2907 struct sk_buff *buf;
2908 struct tlv_desc *rep_tlv;
2913 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2914 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2916 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2920 rep_tlv = (struct tlv_desc *)buf->data;
2921 pb = TLV_DATA(rep_tlv);
2922 pb_len = ULTRA_STRING_MAX_LEN;
2923 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2927 return tipc_cfg_reply_error_string("link not found");
2929 str_len += 1; /* for "\0" */
2930 skb_put(buf, TLV_SPACE(str_len));
2931 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2937 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2938 * @dest: network address of destination node
2939 * @selector: used to select from set of active links
2941 * If no active link can be found, uses default maximum packet size.
2943 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2945 struct tipc_node *n_ptr;
2946 struct tipc_link *l_ptr;
2947 u32 res = MAX_PKT_DEFAULT;
2949 if (dest == tipc_own_addr)
2950 return MAX_MSG_SIZE;
2952 read_lock_bh(&tipc_net_lock);
2953 n_ptr = tipc_node_find(dest);
2955 tipc_node_lock(n_ptr);
2956 l_ptr = n_ptr->active_links[selector & 1];
2958 res = l_ptr->max_pkt;
2959 tipc_node_unlock(n_ptr);
2961 read_unlock_bh(&tipc_net_lock);
2965 static void link_print(struct tipc_link *l_ptr, const char *str)
2967 pr_info("%s Link %x<%s>:", str, l_ptr->addr, l_ptr->b_ptr->name);
2969 if (link_working_unknown(l_ptr))
2971 else if (link_reset_reset(l_ptr))
2973 else if (link_reset_unknown(l_ptr))
2975 else if (link_working_working(l_ptr))