2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, 2012-2014, Ericsson AB
5 * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
40 #include "name_distr.h"
44 #include <linux/pkt_sched.h>
47 * Error message prefixes
49 static const char *link_co_err = "Link changeover error, ";
50 static const char *link_rst_msg = "Resetting link ";
51 static const char *link_unk_evt = "Unknown link event ";
54 * Out-of-range value for link session numbers
56 #define INVALID_SESSION 0x10000
61 #define STARTING_EVT 856384768 /* link processing trigger */
62 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
63 #define TIMEOUT_EVT 560817u /* link timer expired */
66 * The following two 'message types' is really just implementation
67 * data conveniently stored in the message header.
68 * They must not be considered part of the protocol
74 * State value stored in 'exp_msg_count'
76 #define START_CHANGEOVER 100000u
78 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
80 static void link_recv_proto_msg(struct tipc_link *l_ptr, struct sk_buff *buf);
81 static int tipc_link_tunnel_rcv(struct tipc_link **l_ptr,
82 struct sk_buff **buf);
83 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
84 static int link_send_sections_long(struct tipc_port *sender,
85 struct iovec const *msg_sect,
86 unsigned int len, u32 destnode);
87 static void link_state_event(struct tipc_link *l_ptr, u32 event);
88 static void link_reset_statistics(struct tipc_link *l_ptr);
89 static void link_print(struct tipc_link *l_ptr, const char *str);
90 static void link_start(struct tipc_link *l_ptr);
91 static int link_send_long_buf(struct tipc_link *l_ptr, struct sk_buff *buf);
92 static void tipc_link_send_sync(struct tipc_link *l);
93 static void tipc_link_recv_sync(struct tipc_node *n, struct sk_buff *buf);
96 * Simple link routines
98 static unsigned int align(unsigned int i)
100 return (i + 3) & ~3u;
103 static void link_init_max_pkt(struct tipc_link *l_ptr)
107 max_pkt = (l_ptr->b_ptr->mtu & ~3);
108 if (max_pkt > MAX_MSG_SIZE)
109 max_pkt = MAX_MSG_SIZE;
111 l_ptr->max_pkt_target = max_pkt;
112 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
113 l_ptr->max_pkt = l_ptr->max_pkt_target;
115 l_ptr->max_pkt = MAX_PKT_DEFAULT;
117 l_ptr->max_pkt_probes = 0;
120 static u32 link_next_sent(struct tipc_link *l_ptr)
123 return buf_seqno(l_ptr->next_out);
124 return mod(l_ptr->next_out_no);
127 static u32 link_last_sent(struct tipc_link *l_ptr)
129 return mod(link_next_sent(l_ptr) - 1);
133 * Simple non-static link routines (i.e. referenced outside this file)
135 int tipc_link_is_up(struct tipc_link *l_ptr)
139 return link_working_working(l_ptr) || link_working_unknown(l_ptr);
142 int tipc_link_is_active(struct tipc_link *l_ptr)
144 return (l_ptr->owner->active_links[0] == l_ptr) ||
145 (l_ptr->owner->active_links[1] == l_ptr);
149 * link_timeout - handle expiration of link timer
150 * @l_ptr: pointer to link
152 * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
153 * with tipc_link_delete(). (There is no risk that the node will be deleted by
154 * another thread because tipc_link_delete() always cancels the link timer before
155 * tipc_node_delete() is called.)
157 static void link_timeout(struct tipc_link *l_ptr)
159 tipc_node_lock(l_ptr->owner);
161 /* update counters used in statistical profiling of send traffic */
162 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
163 l_ptr->stats.queue_sz_counts++;
165 if (l_ptr->first_out) {
166 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
167 u32 length = msg_size(msg);
169 if ((msg_user(msg) == MSG_FRAGMENTER) &&
170 (msg_type(msg) == FIRST_FRAGMENT)) {
171 length = msg_size(msg_get_wrapped(msg));
174 l_ptr->stats.msg_lengths_total += length;
175 l_ptr->stats.msg_length_counts++;
177 l_ptr->stats.msg_length_profile[0]++;
178 else if (length <= 256)
179 l_ptr->stats.msg_length_profile[1]++;
180 else if (length <= 1024)
181 l_ptr->stats.msg_length_profile[2]++;
182 else if (length <= 4096)
183 l_ptr->stats.msg_length_profile[3]++;
184 else if (length <= 16384)
185 l_ptr->stats.msg_length_profile[4]++;
186 else if (length <= 32768)
187 l_ptr->stats.msg_length_profile[5]++;
189 l_ptr->stats.msg_length_profile[6]++;
193 /* do all other link processing performed on a periodic basis */
195 link_state_event(l_ptr, TIMEOUT_EVT);
198 tipc_link_push_queue(l_ptr);
200 tipc_node_unlock(l_ptr->owner);
203 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
205 k_start_timer(&l_ptr->timer, time);
209 * tipc_link_create - create a new link
210 * @n_ptr: pointer to associated node
211 * @b_ptr: pointer to associated bearer
212 * @media_addr: media address to use when sending messages over link
214 * Returns pointer to link.
216 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
217 struct tipc_bearer *b_ptr,
218 const struct tipc_media_addr *media_addr)
220 struct tipc_link *l_ptr;
221 struct tipc_msg *msg;
223 char addr_string[16];
224 u32 peer = n_ptr->addr;
226 if (n_ptr->link_cnt >= 2) {
227 tipc_addr_string_fill(addr_string, n_ptr->addr);
228 pr_err("Attempt to establish third link to %s\n", addr_string);
232 if (n_ptr->links[b_ptr->identity]) {
233 tipc_addr_string_fill(addr_string, n_ptr->addr);
234 pr_err("Attempt to establish second link on <%s> to %s\n",
235 b_ptr->name, addr_string);
239 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
241 pr_warn("Link creation failed, no memory\n");
246 if_name = strchr(b_ptr->name, ':') + 1;
247 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
248 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
249 tipc_node(tipc_own_addr),
251 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
252 /* note: peer i/f name is updated by reset/activate message */
253 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
254 l_ptr->owner = n_ptr;
255 l_ptr->checkpoint = 1;
256 l_ptr->peer_session = INVALID_SESSION;
257 l_ptr->b_ptr = b_ptr;
258 link_set_supervision_props(l_ptr, b_ptr->tolerance);
259 l_ptr->state = RESET_UNKNOWN;
261 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
263 tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
264 msg_set_size(msg, sizeof(l_ptr->proto_msg));
265 msg_set_session(msg, (tipc_random & 0xffff));
266 msg_set_bearer_id(msg, b_ptr->identity);
267 strcpy((char *)msg_data(msg), if_name);
269 l_ptr->priority = b_ptr->priority;
270 tipc_link_set_queue_limits(l_ptr, b_ptr->window);
272 link_init_max_pkt(l_ptr);
274 l_ptr->next_out_no = 1;
275 INIT_LIST_HEAD(&l_ptr->waiting_ports);
277 link_reset_statistics(l_ptr);
279 tipc_node_attach_link(n_ptr, l_ptr);
281 k_init_timer(&l_ptr->timer, (Handler)link_timeout,
282 (unsigned long)l_ptr);
283 list_add_tail(&l_ptr->link_list, &b_ptr->links);
284 tipc_k_signal((Handler)link_start, (unsigned long)l_ptr);
290 * tipc_link_delete - delete a link
291 * @l_ptr: pointer to link
293 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
294 * This routine must not grab the node lock until after link timer cancellation
295 * to avoid a potential deadlock situation.
297 void tipc_link_delete(struct tipc_link *l_ptr)
300 pr_err("Attempt to delete non-existent link\n");
304 k_cancel_timer(&l_ptr->timer);
306 tipc_node_lock(l_ptr->owner);
307 tipc_link_reset(l_ptr);
308 tipc_node_detach_link(l_ptr->owner, l_ptr);
309 tipc_link_stop(l_ptr);
310 list_del_init(&l_ptr->link_list);
311 tipc_node_unlock(l_ptr->owner);
312 k_term_timer(&l_ptr->timer);
316 static void link_start(struct tipc_link *l_ptr)
318 tipc_node_lock(l_ptr->owner);
319 link_state_event(l_ptr, STARTING_EVT);
320 tipc_node_unlock(l_ptr->owner);
324 * link_schedule_port - schedule port for deferred sending
325 * @l_ptr: pointer to link
326 * @origport: reference to sending port
327 * @sz: amount of data to be sent
329 * Schedules port for renewed sending of messages after link congestion
332 static int link_schedule_port(struct tipc_link *l_ptr, u32 origport, u32 sz)
334 struct tipc_port *p_ptr;
336 spin_lock_bh(&tipc_port_list_lock);
337 p_ptr = tipc_port_lock(origport);
341 if (!list_empty(&p_ptr->wait_list))
343 p_ptr->congested = 1;
344 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
345 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
346 l_ptr->stats.link_congs++;
348 tipc_port_unlock(p_ptr);
350 spin_unlock_bh(&tipc_port_list_lock);
354 void tipc_link_wakeup_ports(struct tipc_link *l_ptr, int all)
356 struct tipc_port *p_ptr;
357 struct tipc_port *temp_p_ptr;
358 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
364 if (!spin_trylock_bh(&tipc_port_list_lock))
366 if (link_congested(l_ptr))
368 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
372 list_del_init(&p_ptr->wait_list);
373 spin_lock_bh(p_ptr->lock);
374 p_ptr->congested = 0;
375 p_ptr->wakeup(p_ptr);
376 win -= p_ptr->waiting_pkts;
377 spin_unlock_bh(p_ptr->lock);
381 spin_unlock_bh(&tipc_port_list_lock);
385 * link_release_outqueue - purge link's outbound message queue
386 * @l_ptr: pointer to link
388 static void link_release_outqueue(struct tipc_link *l_ptr)
390 kfree_skb_list(l_ptr->first_out);
391 l_ptr->first_out = NULL;
392 l_ptr->out_queue_size = 0;
396 * tipc_link_reset_fragments - purge link's inbound message fragments queue
397 * @l_ptr: pointer to link
399 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
401 kfree_skb(l_ptr->reasm_head);
402 l_ptr->reasm_head = NULL;
403 l_ptr->reasm_tail = NULL;
407 * tipc_link_stop - purge all inbound and outbound messages associated with link
408 * @l_ptr: pointer to link
410 void tipc_link_stop(struct tipc_link *l_ptr)
412 kfree_skb_list(l_ptr->oldest_deferred_in);
413 kfree_skb_list(l_ptr->first_out);
414 tipc_link_reset_fragments(l_ptr);
415 kfree_skb(l_ptr->proto_msg_queue);
416 l_ptr->proto_msg_queue = NULL;
419 void tipc_link_reset(struct tipc_link *l_ptr)
421 u32 prev_state = l_ptr->state;
422 u32 checkpoint = l_ptr->next_in_no;
423 int was_active_link = tipc_link_is_active(l_ptr);
425 msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
427 /* Link is down, accept any session */
428 l_ptr->peer_session = INVALID_SESSION;
430 /* Prepare for max packet size negotiation */
431 link_init_max_pkt(l_ptr);
433 l_ptr->state = RESET_UNKNOWN;
435 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
438 tipc_node_link_down(l_ptr->owner, l_ptr);
439 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
441 if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
442 l_ptr->reset_checkpoint = checkpoint;
443 l_ptr->exp_msg_count = START_CHANGEOVER;
446 /* Clean up all queues: */
447 link_release_outqueue(l_ptr);
448 kfree_skb(l_ptr->proto_msg_queue);
449 l_ptr->proto_msg_queue = NULL;
450 kfree_skb_list(l_ptr->oldest_deferred_in);
451 if (!list_empty(&l_ptr->waiting_ports))
452 tipc_link_wakeup_ports(l_ptr, 1);
454 l_ptr->retransm_queue_head = 0;
455 l_ptr->retransm_queue_size = 0;
456 l_ptr->last_out = NULL;
457 l_ptr->first_out = NULL;
458 l_ptr->next_out = NULL;
459 l_ptr->unacked_window = 0;
460 l_ptr->checkpoint = 1;
461 l_ptr->next_out_no = 1;
462 l_ptr->deferred_inqueue_sz = 0;
463 l_ptr->oldest_deferred_in = NULL;
464 l_ptr->newest_deferred_in = NULL;
465 l_ptr->fsm_msg_cnt = 0;
466 l_ptr->stale_count = 0;
467 link_reset_statistics(l_ptr);
471 static void link_activate(struct tipc_link *l_ptr)
473 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
474 tipc_node_link_up(l_ptr->owner, l_ptr);
475 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
479 * link_state_event - link finite state machine
480 * @l_ptr: pointer to link
481 * @event: state machine event to process
483 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
485 struct tipc_link *other;
486 u32 cont_intv = l_ptr->continuity_interval;
488 if (!l_ptr->started && (event != STARTING_EVT))
489 return; /* Not yet. */
491 /* Check whether changeover is going on */
492 if (l_ptr->exp_msg_count) {
493 if (event == TIMEOUT_EVT)
494 link_set_timer(l_ptr, cont_intv);
498 switch (l_ptr->state) {
499 case WORKING_WORKING:
501 case TRAFFIC_MSG_EVT:
505 if (l_ptr->next_in_no != l_ptr->checkpoint) {
506 l_ptr->checkpoint = l_ptr->next_in_no;
507 if (tipc_bclink_acks_missing(l_ptr->owner)) {
508 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
510 l_ptr->fsm_msg_cnt++;
511 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
512 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
514 l_ptr->fsm_msg_cnt++;
516 link_set_timer(l_ptr, cont_intv);
519 l_ptr->state = WORKING_UNKNOWN;
520 l_ptr->fsm_msg_cnt = 0;
521 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
522 l_ptr->fsm_msg_cnt++;
523 link_set_timer(l_ptr, cont_intv / 4);
526 pr_info("%s<%s>, requested by peer\n", link_rst_msg,
528 tipc_link_reset(l_ptr);
529 l_ptr->state = RESET_RESET;
530 l_ptr->fsm_msg_cnt = 0;
531 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
532 l_ptr->fsm_msg_cnt++;
533 link_set_timer(l_ptr, cont_intv);
536 pr_err("%s%u in WW state\n", link_unk_evt, event);
539 case WORKING_UNKNOWN:
541 case TRAFFIC_MSG_EVT:
543 l_ptr->state = WORKING_WORKING;
544 l_ptr->fsm_msg_cnt = 0;
545 link_set_timer(l_ptr, cont_intv);
548 pr_info("%s<%s>, requested by peer while probing\n",
549 link_rst_msg, l_ptr->name);
550 tipc_link_reset(l_ptr);
551 l_ptr->state = RESET_RESET;
552 l_ptr->fsm_msg_cnt = 0;
553 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
554 l_ptr->fsm_msg_cnt++;
555 link_set_timer(l_ptr, cont_intv);
558 if (l_ptr->next_in_no != l_ptr->checkpoint) {
559 l_ptr->state = WORKING_WORKING;
560 l_ptr->fsm_msg_cnt = 0;
561 l_ptr->checkpoint = l_ptr->next_in_no;
562 if (tipc_bclink_acks_missing(l_ptr->owner)) {
563 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
565 l_ptr->fsm_msg_cnt++;
567 link_set_timer(l_ptr, cont_intv);
568 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
569 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
571 l_ptr->fsm_msg_cnt++;
572 link_set_timer(l_ptr, cont_intv / 4);
573 } else { /* Link has failed */
574 pr_warn("%s<%s>, peer not responding\n",
575 link_rst_msg, l_ptr->name);
576 tipc_link_reset(l_ptr);
577 l_ptr->state = RESET_UNKNOWN;
578 l_ptr->fsm_msg_cnt = 0;
579 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
581 l_ptr->fsm_msg_cnt++;
582 link_set_timer(l_ptr, cont_intv);
586 pr_err("%s%u in WU state\n", link_unk_evt, event);
591 case TRAFFIC_MSG_EVT:
594 other = l_ptr->owner->active_links[0];
595 if (other && link_working_unknown(other))
597 l_ptr->state = WORKING_WORKING;
598 l_ptr->fsm_msg_cnt = 0;
599 link_activate(l_ptr);
600 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
601 l_ptr->fsm_msg_cnt++;
602 if (l_ptr->owner->working_links == 1)
603 tipc_link_send_sync(l_ptr);
604 link_set_timer(l_ptr, cont_intv);
607 l_ptr->state = RESET_RESET;
608 l_ptr->fsm_msg_cnt = 0;
609 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
610 l_ptr->fsm_msg_cnt++;
611 link_set_timer(l_ptr, cont_intv);
617 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
618 l_ptr->fsm_msg_cnt++;
619 link_set_timer(l_ptr, cont_intv);
622 pr_err("%s%u in RU state\n", link_unk_evt, event);
627 case TRAFFIC_MSG_EVT:
629 other = l_ptr->owner->active_links[0];
630 if (other && link_working_unknown(other))
632 l_ptr->state = WORKING_WORKING;
633 l_ptr->fsm_msg_cnt = 0;
634 link_activate(l_ptr);
635 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
636 l_ptr->fsm_msg_cnt++;
637 if (l_ptr->owner->working_links == 1)
638 tipc_link_send_sync(l_ptr);
639 link_set_timer(l_ptr, cont_intv);
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 RR state\n", link_unk_evt, event);
653 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
658 * link_bundle_buf(): Append contents of a buffer to
659 * the tail of an existing one.
661 static int link_bundle_buf(struct tipc_link *l_ptr, struct sk_buff *bundler,
664 struct tipc_msg *bundler_msg = buf_msg(bundler);
665 struct tipc_msg *msg = buf_msg(buf);
666 u32 size = msg_size(msg);
667 u32 bundle_size = msg_size(bundler_msg);
668 u32 to_pos = align(bundle_size);
669 u32 pad = to_pos - bundle_size;
671 if (msg_user(bundler_msg) != MSG_BUNDLER)
673 if (msg_type(bundler_msg) != OPEN_MSG)
675 if (skb_tailroom(bundler) < (pad + size))
677 if (l_ptr->max_pkt < (to_pos + size))
680 skb_put(bundler, pad + size);
681 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
682 msg_set_size(bundler_msg, to_pos + size);
683 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
685 l_ptr->stats.sent_bundled++;
689 static void link_add_to_outqueue(struct tipc_link *l_ptr,
691 struct tipc_msg *msg)
693 u32 ack = mod(l_ptr->next_in_no - 1);
694 u32 seqno = mod(l_ptr->next_out_no++);
696 msg_set_word(msg, 2, ((ack << 16) | seqno));
697 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
699 if (l_ptr->first_out) {
700 l_ptr->last_out->next = buf;
701 l_ptr->last_out = buf;
703 l_ptr->first_out = l_ptr->last_out = buf;
705 l_ptr->out_queue_size++;
706 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
707 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
710 static void link_add_chain_to_outqueue(struct tipc_link *l_ptr,
711 struct sk_buff *buf_chain,
715 struct tipc_msg *msg;
717 if (!l_ptr->next_out)
718 l_ptr->next_out = buf_chain;
721 buf_chain = buf_chain->next;
724 msg_set_long_msgno(msg, long_msgno);
725 link_add_to_outqueue(l_ptr, buf, msg);
730 * tipc_link_send_buf() is the 'full path' for messages, called from
731 * inside TIPC when the 'fast path' in tipc_send_buf
732 * has failed, and from link_send()
734 int tipc_link_send_buf(struct tipc_link *l_ptr, struct sk_buff *buf)
736 struct tipc_msg *msg = buf_msg(buf);
737 u32 size = msg_size(msg);
738 u32 dsz = msg_data_sz(msg);
739 u32 queue_size = l_ptr->out_queue_size;
740 u32 imp = tipc_msg_tot_importance(msg);
741 u32 queue_limit = l_ptr->queue_limit[imp];
742 u32 max_packet = l_ptr->max_pkt;
744 /* Match msg importance against queue limits: */
745 if (unlikely(queue_size >= queue_limit)) {
746 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
747 link_schedule_port(l_ptr, msg_origport(msg), size);
752 if (imp > CONN_MANAGER) {
753 pr_warn("%s<%s>, send queue full", link_rst_msg,
755 tipc_link_reset(l_ptr);
760 /* Fragmentation needed ? */
761 if (size > max_packet)
762 return link_send_long_buf(l_ptr, buf);
764 /* Packet can be queued or sent. */
765 if (likely(!link_congested(l_ptr))) {
766 link_add_to_outqueue(l_ptr, buf, msg);
768 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
769 l_ptr->unacked_window = 0;
772 /* Congestion: can message be bundled ? */
773 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
774 (msg_user(msg) != MSG_FRAGMENTER)) {
776 /* Try adding message to an existing bundle */
777 if (l_ptr->next_out &&
778 link_bundle_buf(l_ptr, l_ptr->last_out, buf))
781 /* Try creating a new bundle */
782 if (size <= max_packet * 2 / 3) {
783 struct sk_buff *bundler = tipc_buf_acquire(max_packet);
784 struct tipc_msg bundler_hdr;
787 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
788 INT_H_SIZE, l_ptr->addr);
789 skb_copy_to_linear_data(bundler, &bundler_hdr,
791 skb_trim(bundler, INT_H_SIZE);
792 link_bundle_buf(l_ptr, bundler, buf);
795 l_ptr->stats.sent_bundles++;
799 if (!l_ptr->next_out)
800 l_ptr->next_out = buf;
801 link_add_to_outqueue(l_ptr, buf, msg);
806 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
807 * not been selected yet, and the the owner node is not locked
808 * Called by TIPC internal users, e.g. the name distributor
810 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
812 struct tipc_link *l_ptr;
813 struct tipc_node *n_ptr;
814 int res = -ELINKCONG;
816 read_lock_bh(&tipc_net_lock);
817 n_ptr = tipc_node_find(dest);
819 tipc_node_lock(n_ptr);
820 l_ptr = n_ptr->active_links[selector & 1];
822 res = tipc_link_send_buf(l_ptr, buf);
825 tipc_node_unlock(n_ptr);
829 read_unlock_bh(&tipc_net_lock);
834 * tipc_link_send_sync - synchronize broadcast link endpoints.
836 * Give a newly added peer node the sequence number where it should
837 * start receiving and acking broadcast packets.
839 * Called with node locked
841 static void tipc_link_send_sync(struct tipc_link *l)
844 struct tipc_msg *msg;
846 buf = tipc_buf_acquire(INT_H_SIZE);
851 tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, l->addr);
852 msg_set_last_bcast(msg, l->owner->bclink.acked);
853 link_add_chain_to_outqueue(l, buf, 0);
854 tipc_link_push_queue(l);
858 * tipc_link_recv_sync - synchronize broadcast link endpoints.
859 * Receive the sequence number where we should start receiving and
860 * acking broadcast packets from a newly added peer node, and open
861 * up for reception of such packets.
863 * Called with node locked
865 static void tipc_link_recv_sync(struct tipc_node *n, struct sk_buff *buf)
867 struct tipc_msg *msg = buf_msg(buf);
869 n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
870 n->bclink.recv_permitted = true;
875 * tipc_link_send_names - send name table entries to new neighbor
877 * Send routine for bulk delivery of name table messages when contact
878 * with a new neighbor occurs. No link congestion checking is performed
879 * because name table messages *must* be delivered. The messages must be
880 * small enough not to require fragmentation.
881 * Called without any locks held.
883 void tipc_link_send_names(struct list_head *message_list, u32 dest)
885 struct tipc_node *n_ptr;
886 struct tipc_link *l_ptr;
888 struct sk_buff *temp_buf;
890 if (list_empty(message_list))
893 read_lock_bh(&tipc_net_lock);
894 n_ptr = tipc_node_find(dest);
896 tipc_node_lock(n_ptr);
897 l_ptr = n_ptr->active_links[0];
899 /* convert circular list to linear list */
900 ((struct sk_buff *)message_list->prev)->next = NULL;
901 link_add_chain_to_outqueue(l_ptr,
902 (struct sk_buff *)message_list->next, 0);
903 tipc_link_push_queue(l_ptr);
904 INIT_LIST_HEAD(message_list);
906 tipc_node_unlock(n_ptr);
908 read_unlock_bh(&tipc_net_lock);
910 /* discard the messages if they couldn't be sent */
911 list_for_each_safe(buf, temp_buf, ((struct sk_buff *)message_list)) {
912 list_del((struct list_head *)buf);
918 * link_send_buf_fast: Entry for data messages where the
919 * destination link is known and the header is complete,
920 * inclusive total message length. Very time critical.
921 * Link is locked. Returns user data length.
923 static int link_send_buf_fast(struct tipc_link *l_ptr, struct sk_buff *buf,
926 struct tipc_msg *msg = buf_msg(buf);
927 int res = msg_data_sz(msg);
929 if (likely(!link_congested(l_ptr))) {
930 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
931 link_add_to_outqueue(l_ptr, buf, msg);
932 tipc_bearer_send(l_ptr->b_ptr, buf,
934 l_ptr->unacked_window = 0;
938 *used_max_pkt = l_ptr->max_pkt;
940 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
944 * tipc_link_send_sections_fast: Entry for messages where the
945 * destination processor is known and the header is complete,
946 * except for total message length.
947 * Returns user data length or errno.
949 int tipc_link_send_sections_fast(struct tipc_port *sender,
950 struct iovec const *msg_sect,
951 unsigned int len, u32 destaddr)
953 struct tipc_msg *hdr = &sender->phdr;
954 struct tipc_link *l_ptr;
956 struct tipc_node *node;
958 u32 selector = msg_origport(hdr) & 1;
962 * Try building message using port's max_pkt hint.
963 * (Must not hold any locks while building message.)
965 res = tipc_msg_build(hdr, msg_sect, len, sender->max_pkt, &buf);
966 /* Exit if build request was invalid */
967 if (unlikely(res < 0))
970 read_lock_bh(&tipc_net_lock);
971 node = tipc_node_find(destaddr);
973 tipc_node_lock(node);
974 l_ptr = node->active_links[selector];
977 res = link_send_buf_fast(l_ptr, buf,
980 tipc_node_unlock(node);
981 read_unlock_bh(&tipc_net_lock);
985 /* Exit if link (or bearer) is congested */
986 if (link_congested(l_ptr)) {
987 res = link_schedule_port(l_ptr,
993 * Message size exceeds max_pkt hint; update hint,
994 * then re-try fast path or fragment the message
996 sender->max_pkt = l_ptr->max_pkt;
997 tipc_node_unlock(node);
998 read_unlock_bh(&tipc_net_lock);
1001 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1004 return link_send_sections_long(sender, msg_sect, len,
1007 tipc_node_unlock(node);
1009 read_unlock_bh(&tipc_net_lock);
1011 /* Couldn't find a link to the destination node */
1013 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1015 return tipc_port_reject_sections(sender, hdr, msg_sect,
1016 len, TIPC_ERR_NO_NODE);
1021 * link_send_sections_long(): Entry for long messages where the
1022 * destination node is known and the header is complete,
1023 * inclusive total message length.
1024 * Link and bearer congestion status have been checked to be ok,
1025 * and are ignored if they change.
1027 * Note that fragments do not use the full link MTU so that they won't have
1028 * to undergo refragmentation if link changeover causes them to be sent
1029 * over another link with an additional tunnel header added as prefix.
1030 * (Refragmentation will still occur if the other link has a smaller MTU.)
1032 * Returns user data length or errno.
1034 static int link_send_sections_long(struct tipc_port *sender,
1035 struct iovec const *msg_sect,
1036 unsigned int len, u32 destaddr)
1038 struct tipc_link *l_ptr;
1039 struct tipc_node *node;
1040 struct tipc_msg *hdr = &sender->phdr;
1042 u32 max_pkt, fragm_sz, rest;
1043 struct tipc_msg fragm_hdr;
1044 struct sk_buff *buf, *buf_chain, *prev;
1045 u32 fragm_crs, fragm_rest, hsz, sect_rest;
1046 const unchar __user *sect_crs;
1053 max_pkt = sender->max_pkt - INT_H_SIZE;
1054 /* leave room for tunnel header in case of link changeover */
1055 fragm_sz = max_pkt - INT_H_SIZE;
1056 /* leave room for fragmentation header in each fragment */
1064 /* Prepare reusable fragment header */
1065 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1066 INT_H_SIZE, msg_destnode(hdr));
1067 msg_set_size(&fragm_hdr, max_pkt);
1068 msg_set_fragm_no(&fragm_hdr, 1);
1070 /* Prepare header of first fragment */
1071 buf_chain = buf = tipc_buf_acquire(max_pkt);
1075 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1076 hsz = msg_hdr_sz(hdr);
1077 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1079 /* Chop up message */
1080 fragm_crs = INT_H_SIZE + hsz;
1081 fragm_rest = fragm_sz - hsz;
1083 do { /* For all sections */
1087 sect_rest = msg_sect[++curr_sect].iov_len;
1088 sect_crs = msg_sect[curr_sect].iov_base;
1091 if (sect_rest < fragm_rest)
1096 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1099 kfree_skb_list(buf_chain);
1108 if (!fragm_rest && rest) {
1110 /* Initiate new fragment: */
1111 if (rest <= fragm_sz) {
1113 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1115 msg_set_type(&fragm_hdr, FRAGMENT);
1117 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1118 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1120 buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1128 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1129 fragm_crs = INT_H_SIZE;
1130 fragm_rest = fragm_sz;
1135 * Now we have a buffer chain. Select a link and check
1136 * that packet size is still OK
1138 node = tipc_node_find(destaddr);
1140 tipc_node_lock(node);
1141 l_ptr = node->active_links[sender->ref & 1];
1143 tipc_node_unlock(node);
1146 if (l_ptr->max_pkt < max_pkt) {
1147 sender->max_pkt = l_ptr->max_pkt;
1148 tipc_node_unlock(node);
1149 kfree_skb_list(buf_chain);
1154 kfree_skb_list(buf_chain);
1155 return tipc_port_reject_sections(sender, hdr, msg_sect,
1156 len, TIPC_ERR_NO_NODE);
1159 /* Append chain of fragments to send queue & send them */
1160 l_ptr->long_msg_seq_no++;
1161 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1162 l_ptr->stats.sent_fragments += fragm_no;
1163 l_ptr->stats.sent_fragmented++;
1164 tipc_link_push_queue(l_ptr);
1165 tipc_node_unlock(node);
1170 * tipc_link_push_packet: Push one unsent packet to the media
1172 static u32 tipc_link_push_packet(struct tipc_link *l_ptr)
1174 struct sk_buff *buf = l_ptr->first_out;
1175 u32 r_q_size = l_ptr->retransm_queue_size;
1176 u32 r_q_head = l_ptr->retransm_queue_head;
1178 /* Step to position where retransmission failed, if any, */
1179 /* consider that buffers may have been released in meantime */
1180 if (r_q_size && buf) {
1181 u32 last = lesser(mod(r_q_head + r_q_size),
1182 link_last_sent(l_ptr));
1183 u32 first = buf_seqno(buf);
1185 while (buf && less(first, r_q_head)) {
1186 first = mod(first + 1);
1189 l_ptr->retransm_queue_head = r_q_head = first;
1190 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1193 /* Continue retransmission now, if there is anything: */
1194 if (r_q_size && buf) {
1195 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1196 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1197 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1198 l_ptr->retransm_queue_head = mod(++r_q_head);
1199 l_ptr->retransm_queue_size = --r_q_size;
1200 l_ptr->stats.retransmitted++;
1204 /* Send deferred protocol message, if any: */
1205 buf = l_ptr->proto_msg_queue;
1207 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1208 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1209 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1210 l_ptr->unacked_window = 0;
1212 l_ptr->proto_msg_queue = NULL;
1216 /* Send one deferred data message, if send window not full: */
1217 buf = l_ptr->next_out;
1219 struct tipc_msg *msg = buf_msg(buf);
1220 u32 next = msg_seqno(msg);
1221 u32 first = buf_seqno(l_ptr->first_out);
1223 if (mod(next - first) < l_ptr->queue_limit[0]) {
1224 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1225 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1226 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1227 if (msg_user(msg) == MSG_BUNDLER)
1228 msg_set_type(msg, CLOSED_MSG);
1229 l_ptr->next_out = buf->next;
1237 * push_queue(): push out the unsent messages of a link where
1238 * congestion has abated. Node is locked
1240 void tipc_link_push_queue(struct tipc_link *l_ptr)
1245 res = tipc_link_push_packet(l_ptr);
1249 static void link_reset_all(unsigned long addr)
1251 struct tipc_node *n_ptr;
1252 char addr_string[16];
1255 read_lock_bh(&tipc_net_lock);
1256 n_ptr = tipc_node_find((u32)addr);
1258 read_unlock_bh(&tipc_net_lock);
1259 return; /* node no longer exists */
1262 tipc_node_lock(n_ptr);
1264 pr_warn("Resetting all links to %s\n",
1265 tipc_addr_string_fill(addr_string, n_ptr->addr));
1267 for (i = 0; i < MAX_BEARERS; i++) {
1268 if (n_ptr->links[i]) {
1269 link_print(n_ptr->links[i], "Resetting link\n");
1270 tipc_link_reset(n_ptr->links[i]);
1274 tipc_node_unlock(n_ptr);
1275 read_unlock_bh(&tipc_net_lock);
1278 static void link_retransmit_failure(struct tipc_link *l_ptr,
1279 struct sk_buff *buf)
1281 struct tipc_msg *msg = buf_msg(buf);
1283 pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
1286 /* Handle failure on standard link */
1287 link_print(l_ptr, "Resetting link\n");
1288 tipc_link_reset(l_ptr);
1291 /* Handle failure on broadcast link */
1292 struct tipc_node *n_ptr;
1293 char addr_string[16];
1295 pr_info("Msg seq number: %u, ", msg_seqno(msg));
1296 pr_cont("Outstanding acks: %lu\n",
1297 (unsigned long) TIPC_SKB_CB(buf)->handle);
1299 n_ptr = tipc_bclink_retransmit_to();
1300 tipc_node_lock(n_ptr);
1302 tipc_addr_string_fill(addr_string, n_ptr->addr);
1303 pr_info("Broadcast link info for %s\n", addr_string);
1304 pr_info("Reception permitted: %d, Acked: %u\n",
1305 n_ptr->bclink.recv_permitted,
1306 n_ptr->bclink.acked);
1307 pr_info("Last in: %u, Oos state: %u, Last sent: %u\n",
1308 n_ptr->bclink.last_in,
1309 n_ptr->bclink.oos_state,
1310 n_ptr->bclink.last_sent);
1312 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1314 tipc_node_unlock(n_ptr);
1316 l_ptr->stale_count = 0;
1320 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
1323 struct tipc_msg *msg;
1330 /* Detect repeated retransmit failures */
1331 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1332 if (++l_ptr->stale_count > 100) {
1333 link_retransmit_failure(l_ptr, buf);
1337 l_ptr->last_retransmitted = msg_seqno(msg);
1338 l_ptr->stale_count = 1;
1341 while (retransmits && (buf != l_ptr->next_out) && buf) {
1343 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1344 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1345 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1348 l_ptr->stats.retransmitted++;
1351 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1355 * link_insert_deferred_queue - insert deferred messages back into receive chain
1357 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1358 struct sk_buff *buf)
1362 if (l_ptr->oldest_deferred_in == NULL)
1365 seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1366 if (seq_no == mod(l_ptr->next_in_no)) {
1367 l_ptr->newest_deferred_in->next = buf;
1368 buf = l_ptr->oldest_deferred_in;
1369 l_ptr->oldest_deferred_in = NULL;
1370 l_ptr->deferred_inqueue_sz = 0;
1376 * link_recv_buf_validate - validate basic format of received message
1378 * This routine ensures a TIPC message has an acceptable header, and at least
1379 * as much data as the header indicates it should. The routine also ensures
1380 * that the entire message header is stored in the main fragment of the message
1381 * buffer, to simplify future access to message header fields.
1383 * Note: Having extra info present in the message header or data areas is OK.
1384 * TIPC will ignore the excess, under the assumption that it is optional info
1385 * introduced by a later release of the protocol.
1387 static int link_recv_buf_validate(struct sk_buff *buf)
1389 static u32 min_data_hdr_size[8] = {
1390 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1391 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1394 struct tipc_msg *msg;
1400 if (unlikely(buf->len < MIN_H_SIZE))
1403 msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1407 if (unlikely(msg_version(msg) != TIPC_VERSION))
1410 size = msg_size(msg);
1411 hdr_size = msg_hdr_sz(msg);
1412 min_hdr_size = msg_isdata(msg) ?
1413 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1415 if (unlikely((hdr_size < min_hdr_size) ||
1416 (size < hdr_size) ||
1417 (buf->len < size) ||
1418 (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1421 return pskb_may_pull(buf, hdr_size);
1425 * tipc_rcv - process TIPC packets/messages arriving from off-node
1426 * @head: pointer to message buffer chain
1427 * @tb_ptr: pointer to bearer message arrived on
1429 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1430 * structure (i.e. cannot be NULL), but bearer can be inactive.
1432 void tipc_rcv(struct sk_buff *head, struct tipc_bearer *b_ptr)
1434 read_lock_bh(&tipc_net_lock);
1436 struct tipc_node *n_ptr;
1437 struct tipc_link *l_ptr;
1438 struct sk_buff *crs;
1439 struct sk_buff *buf = head;
1440 struct tipc_msg *msg;
1448 /* Ensure bearer is still enabled */
1449 if (unlikely(!b_ptr->active))
1452 /* Ensure message is well-formed */
1453 if (unlikely(!link_recv_buf_validate(buf)))
1456 /* Ensure message data is a single contiguous unit */
1457 if (unlikely(skb_linearize(buf)))
1460 /* Handle arrival of a non-unicast link message */
1463 if (unlikely(msg_non_seq(msg))) {
1464 if (msg_user(msg) == LINK_CONFIG)
1465 tipc_disc_recv_msg(buf, b_ptr);
1467 tipc_bclink_recv_pkt(buf);
1471 /* Discard unicast link messages destined for another node */
1472 if (unlikely(!msg_short(msg) &&
1473 (msg_destnode(msg) != tipc_own_addr)))
1476 /* Locate neighboring node that sent message */
1477 n_ptr = tipc_node_find(msg_prevnode(msg));
1478 if (unlikely(!n_ptr))
1480 tipc_node_lock(n_ptr);
1482 /* Locate unicast link endpoint that should handle message */
1483 l_ptr = n_ptr->links[b_ptr->identity];
1484 if (unlikely(!l_ptr))
1485 goto unlock_discard;
1487 /* Verify that communication with node is currently allowed */
1488 if ((n_ptr->block_setup & WAIT_PEER_DOWN) &&
1489 msg_user(msg) == LINK_PROTOCOL &&
1490 (msg_type(msg) == RESET_MSG ||
1491 msg_type(msg) == ACTIVATE_MSG) &&
1492 !msg_redundant_link(msg))
1493 n_ptr->block_setup &= ~WAIT_PEER_DOWN;
1495 if (n_ptr->block_setup)
1496 goto unlock_discard;
1498 /* Validate message sequence number info */
1499 seq_no = msg_seqno(msg);
1500 ackd = msg_ack(msg);
1502 /* Release acked messages */
1503 if (n_ptr->bclink.recv_permitted)
1504 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1506 crs = l_ptr->first_out;
1507 while ((crs != l_ptr->next_out) &&
1508 less_eq(buf_seqno(crs), ackd)) {
1509 struct sk_buff *next = crs->next;
1516 l_ptr->first_out = crs;
1517 l_ptr->out_queue_size -= released;
1520 /* Try sending any messages link endpoint has pending */
1521 if (unlikely(l_ptr->next_out))
1522 tipc_link_push_queue(l_ptr);
1523 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1524 tipc_link_wakeup_ports(l_ptr, 0);
1525 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1526 l_ptr->stats.sent_acks++;
1527 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1530 /* Now (finally!) process the incoming message */
1532 if (unlikely(!link_working_working(l_ptr))) {
1533 if (msg_user(msg) == LINK_PROTOCOL) {
1534 link_recv_proto_msg(l_ptr, buf);
1535 head = link_insert_deferred_queue(l_ptr, head);
1536 tipc_node_unlock(n_ptr);
1540 /* Traffic message. Conditionally activate link */
1541 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1543 if (link_working_working(l_ptr)) {
1544 /* Re-insert buffer in front of queue */
1547 tipc_node_unlock(n_ptr);
1550 goto unlock_discard;
1553 /* Link is now in state WORKING_WORKING */
1554 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1555 link_handle_out_of_seq_msg(l_ptr, buf);
1556 head = link_insert_deferred_queue(l_ptr, head);
1557 tipc_node_unlock(n_ptr);
1560 l_ptr->next_in_no++;
1561 if (unlikely(l_ptr->oldest_deferred_in))
1562 head = link_insert_deferred_queue(l_ptr, head);
1564 if (likely(msg_isdata(msg))) {
1565 tipc_node_unlock(n_ptr);
1566 tipc_port_recv_msg(buf);
1569 switch (msg_user(msg)) {
1572 l_ptr->stats.recv_bundles++;
1573 l_ptr->stats.recv_bundled += msg_msgcnt(msg);
1574 tipc_node_unlock(n_ptr);
1575 tipc_link_recv_bundle(buf);
1577 case NAME_DISTRIBUTOR:
1578 n_ptr->bclink.recv_permitted = true;
1579 tipc_node_unlock(n_ptr);
1580 tipc_named_recv(buf);
1582 case BCAST_PROTOCOL:
1583 tipc_link_recv_sync(n_ptr, buf);
1584 tipc_node_unlock(n_ptr);
1587 tipc_node_unlock(n_ptr);
1588 tipc_port_recv_proto_msg(buf);
1590 case MSG_FRAGMENTER:
1591 l_ptr->stats.recv_fragments++;
1592 ret = tipc_link_recv_fragment(&l_ptr->reasm_head,
1595 if (ret == LINK_REASM_COMPLETE) {
1596 l_ptr->stats.recv_fragmented++;
1600 if (ret == LINK_REASM_ERROR)
1601 tipc_link_reset(l_ptr);
1602 tipc_node_unlock(n_ptr);
1604 case CHANGEOVER_PROTOCOL:
1605 type = msg_type(msg);
1606 if (tipc_link_tunnel_rcv(&l_ptr, &buf)) {
1608 seq_no = msg_seqno(msg);
1609 if (type == ORIGINAL_MSG)
1611 goto protocol_check;
1619 tipc_node_unlock(n_ptr);
1620 tipc_net_route_msg(buf);
1624 tipc_node_unlock(n_ptr);
1628 read_unlock_bh(&tipc_net_lock);
1632 * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1634 * Returns increase in queue length (i.e. 0 or 1)
1636 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1637 struct sk_buff *buf)
1639 struct sk_buff *queue_buf;
1640 struct sk_buff **prev;
1641 u32 seq_no = buf_seqno(buf);
1646 if (*head == NULL) {
1647 *head = *tail = buf;
1652 if (less(buf_seqno(*tail), seq_no)) {
1653 (*tail)->next = buf;
1658 /* Locate insertion point in queue, then insert; discard if duplicate */
1662 u32 curr_seqno = buf_seqno(queue_buf);
1664 if (seq_no == curr_seqno) {
1669 if (less(seq_no, curr_seqno))
1672 prev = &queue_buf->next;
1673 queue_buf = queue_buf->next;
1676 buf->next = queue_buf;
1682 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1684 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1685 struct sk_buff *buf)
1687 u32 seq_no = buf_seqno(buf);
1689 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1690 link_recv_proto_msg(l_ptr, buf);
1694 /* Record OOS packet arrival (force mismatch on next timeout) */
1695 l_ptr->checkpoint--;
1698 * Discard packet if a duplicate; otherwise add it to deferred queue
1699 * and notify peer of gap as per protocol specification
1701 if (less(seq_no, mod(l_ptr->next_in_no))) {
1702 l_ptr->stats.duplicates++;
1707 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1708 &l_ptr->newest_deferred_in, buf)) {
1709 l_ptr->deferred_inqueue_sz++;
1710 l_ptr->stats.deferred_recv++;
1711 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1712 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1714 l_ptr->stats.duplicates++;
1718 * Send protocol message to the other endpoint.
1720 void tipc_link_send_proto_msg(struct tipc_link *l_ptr, u32 msg_typ,
1721 int probe_msg, u32 gap, u32 tolerance,
1722 u32 priority, u32 ack_mtu)
1724 struct sk_buff *buf = NULL;
1725 struct tipc_msg *msg = l_ptr->pmsg;
1726 u32 msg_size = sizeof(l_ptr->proto_msg);
1729 /* Discard any previous message that was deferred due to congestion */
1730 if (l_ptr->proto_msg_queue) {
1731 kfree_skb(l_ptr->proto_msg_queue);
1732 l_ptr->proto_msg_queue = NULL;
1735 /* Don't send protocol message during link changeover */
1736 if (l_ptr->exp_msg_count)
1739 /* Abort non-RESET send if communication with node is prohibited */
1740 if ((l_ptr->owner->block_setup) && (msg_typ != RESET_MSG))
1743 /* Create protocol message with "out-of-sequence" sequence number */
1744 msg_set_type(msg, msg_typ);
1745 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
1746 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1747 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1749 if (msg_typ == STATE_MSG) {
1750 u32 next_sent = mod(l_ptr->next_out_no);
1752 if (!tipc_link_is_up(l_ptr))
1754 if (l_ptr->next_out)
1755 next_sent = buf_seqno(l_ptr->next_out);
1756 msg_set_next_sent(msg, next_sent);
1757 if (l_ptr->oldest_deferred_in) {
1758 u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1759 gap = mod(rec - mod(l_ptr->next_in_no));
1761 msg_set_seq_gap(msg, gap);
1763 l_ptr->stats.sent_nacks++;
1764 msg_set_link_tolerance(msg, tolerance);
1765 msg_set_linkprio(msg, priority);
1766 msg_set_max_pkt(msg, ack_mtu);
1767 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1768 msg_set_probe(msg, probe_msg != 0);
1770 u32 mtu = l_ptr->max_pkt;
1772 if ((mtu < l_ptr->max_pkt_target) &&
1773 link_working_working(l_ptr) &&
1774 l_ptr->fsm_msg_cnt) {
1775 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1776 if (l_ptr->max_pkt_probes == 10) {
1777 l_ptr->max_pkt_target = (msg_size - 4);
1778 l_ptr->max_pkt_probes = 0;
1779 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1781 l_ptr->max_pkt_probes++;
1784 l_ptr->stats.sent_probes++;
1786 l_ptr->stats.sent_states++;
1787 } else { /* RESET_MSG or ACTIVATE_MSG */
1788 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1789 msg_set_seq_gap(msg, 0);
1790 msg_set_next_sent(msg, 1);
1791 msg_set_probe(msg, 0);
1792 msg_set_link_tolerance(msg, l_ptr->tolerance);
1793 msg_set_linkprio(msg, l_ptr->priority);
1794 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1797 r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1798 msg_set_redundant_link(msg, r_flag);
1799 msg_set_linkprio(msg, l_ptr->priority);
1800 msg_set_size(msg, msg_size);
1802 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1804 buf = tipc_buf_acquire(msg_size);
1808 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1809 buf->priority = TC_PRIO_CONTROL;
1811 tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr);
1812 l_ptr->unacked_window = 0;
1817 * Receive protocol message :
1818 * Note that network plane id propagates through the network, and may
1819 * change at any time. The node with lowest address rules
1821 static void link_recv_proto_msg(struct tipc_link *l_ptr, struct sk_buff *buf)
1827 struct tipc_msg *msg = buf_msg(buf);
1829 /* Discard protocol message during link changeover */
1830 if (l_ptr->exp_msg_count)
1833 /* record unnumbered packet arrival (force mismatch on next timeout) */
1834 l_ptr->checkpoint--;
1836 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
1837 if (tipc_own_addr > msg_prevnode(msg))
1838 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
1840 switch (msg_type(msg)) {
1843 if (!link_working_unknown(l_ptr) &&
1844 (l_ptr->peer_session != INVALID_SESSION)) {
1845 if (less_eq(msg_session(msg), l_ptr->peer_session))
1846 break; /* duplicate or old reset: ignore */
1849 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1850 link_working_unknown(l_ptr))) {
1852 * peer has lost contact -- don't allow peer's links
1853 * to reactivate before we recognize loss & clean up
1855 l_ptr->owner->block_setup = WAIT_NODE_DOWN;
1858 link_state_event(l_ptr, RESET_MSG);
1862 /* Update link settings according other endpoint's values */
1863 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1865 msg_tol = msg_link_tolerance(msg);
1866 if (msg_tol > l_ptr->tolerance)
1867 link_set_supervision_props(l_ptr, msg_tol);
1869 if (msg_linkprio(msg) > l_ptr->priority)
1870 l_ptr->priority = msg_linkprio(msg);
1872 max_pkt_info = msg_max_pkt(msg);
1874 if (max_pkt_info < l_ptr->max_pkt_target)
1875 l_ptr->max_pkt_target = max_pkt_info;
1876 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1877 l_ptr->max_pkt = l_ptr->max_pkt_target;
1879 l_ptr->max_pkt = l_ptr->max_pkt_target;
1882 /* Synchronize broadcast link info, if not done previously */
1883 if (!tipc_node_is_up(l_ptr->owner)) {
1884 l_ptr->owner->bclink.last_sent =
1885 l_ptr->owner->bclink.last_in =
1886 msg_last_bcast(msg);
1887 l_ptr->owner->bclink.oos_state = 0;
1890 l_ptr->peer_session = msg_session(msg);
1891 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1893 if (msg_type(msg) == ACTIVATE_MSG)
1894 link_state_event(l_ptr, ACTIVATE_MSG);
1898 msg_tol = msg_link_tolerance(msg);
1900 link_set_supervision_props(l_ptr, msg_tol);
1902 if (msg_linkprio(msg) &&
1903 (msg_linkprio(msg) != l_ptr->priority)) {
1904 pr_warn("%s<%s>, priority change %u->%u\n",
1905 link_rst_msg, l_ptr->name, l_ptr->priority,
1907 l_ptr->priority = msg_linkprio(msg);
1908 tipc_link_reset(l_ptr); /* Enforce change to take effect */
1911 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1912 l_ptr->stats.recv_states++;
1913 if (link_reset_unknown(l_ptr))
1916 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1917 rec_gap = mod(msg_next_sent(msg) -
1918 mod(l_ptr->next_in_no));
1921 max_pkt_ack = msg_max_pkt(msg);
1922 if (max_pkt_ack > l_ptr->max_pkt) {
1923 l_ptr->max_pkt = max_pkt_ack;
1924 l_ptr->max_pkt_probes = 0;
1928 if (msg_probe(msg)) {
1929 l_ptr->stats.recv_probes++;
1930 if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1931 max_pkt_ack = msg_size(msg);
1934 /* Protocol message before retransmits, reduce loss risk */
1935 if (l_ptr->owner->bclink.recv_permitted)
1936 tipc_bclink_update_link_state(l_ptr->owner,
1937 msg_last_bcast(msg));
1939 if (rec_gap || (msg_probe(msg))) {
1940 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
1941 0, rec_gap, 0, 0, max_pkt_ack);
1943 if (msg_seq_gap(msg)) {
1944 l_ptr->stats.recv_nacks++;
1945 tipc_link_retransmit(l_ptr, l_ptr->first_out,
1955 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1956 * a different bearer. Owner node is locked.
1958 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1959 struct tipc_msg *tunnel_hdr,
1960 struct tipc_msg *msg,
1963 struct tipc_link *tunnel;
1964 struct sk_buff *buf;
1965 u32 length = msg_size(msg);
1967 tunnel = l_ptr->owner->active_links[selector & 1];
1968 if (!tipc_link_is_up(tunnel)) {
1969 pr_warn("%stunnel link no longer available\n", link_co_err);
1972 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1973 buf = tipc_buf_acquire(length + INT_H_SIZE);
1975 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1978 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
1979 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
1980 tipc_link_send_buf(tunnel, buf);
1984 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1985 * link is still active. We can do failover. Tunnel the failing link's
1986 * whole send queue via the remaining link. This way, we don't lose
1987 * any packets, and sequence order is preserved for subsequent traffic
1988 * sent over the remaining link. Owner node is locked.
1990 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
1992 u32 msgcount = l_ptr->out_queue_size;
1993 struct sk_buff *crs = l_ptr->first_out;
1994 struct tipc_link *tunnel = l_ptr->owner->active_links[0];
1995 struct tipc_msg tunnel_hdr;
2001 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2002 ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2003 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2004 msg_set_msgcnt(&tunnel_hdr, msgcount);
2006 if (!l_ptr->first_out) {
2007 struct sk_buff *buf;
2009 buf = tipc_buf_acquire(INT_H_SIZE);
2011 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2012 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2013 tipc_link_send_buf(tunnel, buf);
2015 pr_warn("%sunable to send changeover msg\n",
2021 split_bundles = (l_ptr->owner->active_links[0] !=
2022 l_ptr->owner->active_links[1]);
2025 struct tipc_msg *msg = buf_msg(crs);
2027 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2028 struct tipc_msg *m = msg_get_wrapped(msg);
2029 unchar *pos = (unchar *)m;
2031 msgcount = msg_msgcnt(msg);
2032 while (msgcount--) {
2033 msg_set_seqno(m, msg_seqno(msg));
2034 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
2035 msg_link_selector(m));
2036 pos += align(msg_size(m));
2037 m = (struct tipc_msg *)pos;
2040 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
2041 msg_link_selector(msg));
2047 /* tipc_link_dup_send_queue(): A second link has become active. Tunnel a
2048 * duplicate of the first link's send queue via the new link. This way, we
2049 * are guaranteed that currently queued packets from a socket are delivered
2050 * before future traffic from the same socket, even if this is using the
2051 * new link. The last arriving copy of each duplicate packet is dropped at
2052 * the receiving end by the regular protocol check, so packet cardinality
2053 * and sequence order is preserved per sender/receiver socket pair.
2054 * Owner node is locked.
2056 void tipc_link_dup_send_queue(struct tipc_link *l_ptr,
2057 struct tipc_link *tunnel)
2059 struct sk_buff *iter;
2060 struct tipc_msg tunnel_hdr;
2062 tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2063 DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2064 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2065 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2066 iter = l_ptr->first_out;
2068 struct sk_buff *outbuf;
2069 struct tipc_msg *msg = buf_msg(iter);
2070 u32 length = msg_size(msg);
2072 if (msg_user(msg) == MSG_BUNDLER)
2073 msg_set_type(msg, CLOSED_MSG);
2074 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2075 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2076 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2077 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2078 if (outbuf == NULL) {
2079 pr_warn("%sunable to send duplicate msg\n",
2083 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2084 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2086 tipc_link_send_buf(tunnel, outbuf);
2087 if (!tipc_link_is_up(l_ptr))
2094 * buf_extract - extracts embedded TIPC message from another message
2095 * @skb: encapsulating message buffer
2096 * @from_pos: offset to extract from
2098 * Returns a new message buffer containing an embedded message. The
2099 * encapsulating message itself is left unchanged.
2101 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2103 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2104 u32 size = msg_size(msg);
2107 eb = tipc_buf_acquire(size);
2109 skb_copy_to_linear_data(eb, msg, size);
2113 /* tipc_link_tunnel_rcv(): Receive a tunneled packet, sent
2114 * via other link as result of a failover (ORIGINAL_MSG) or
2115 * a new active link (DUPLICATE_MSG). Failover packets are
2116 * returned to the active link for delivery upwards.
2117 * Owner node is locked.
2119 static int tipc_link_tunnel_rcv(struct tipc_link **l_ptr,
2120 struct sk_buff **buf)
2122 struct sk_buff *tunnel_buf = *buf;
2123 struct tipc_link *dest_link;
2124 struct tipc_msg *msg;
2125 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2126 u32 msg_typ = msg_type(tunnel_msg);
2127 u32 msg_count = msg_msgcnt(tunnel_msg);
2128 u32 bearer_id = msg_bearer_id(tunnel_msg);
2130 if (bearer_id >= MAX_BEARERS)
2132 dest_link = (*l_ptr)->owner->links[bearer_id];
2135 if (dest_link == *l_ptr) {
2136 pr_err("Unexpected changeover message on link <%s>\n",
2141 msg = msg_get_wrapped(tunnel_msg);
2143 if (msg_typ == DUPLICATE_MSG) {
2144 if (less(msg_seqno(msg), mod(dest_link->next_in_no)))
2146 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2148 pr_warn("%sduplicate msg dropped\n", link_co_err);
2151 kfree_skb(tunnel_buf);
2155 /* First original message ?: */
2156 if (tipc_link_is_up(dest_link)) {
2157 pr_info("%s<%s>, changeover initiated by peer\n", link_rst_msg,
2159 tipc_link_reset(dest_link);
2160 dest_link->exp_msg_count = msg_count;
2163 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2164 dest_link->exp_msg_count = msg_count;
2169 /* Receive original message */
2170 if (dest_link->exp_msg_count == 0) {
2171 pr_warn("%sgot too many tunnelled messages\n", link_co_err);
2174 dest_link->exp_msg_count--;
2175 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2178 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2180 kfree_skb(tunnel_buf);
2183 pr_warn("%soriginal msg dropped\n", link_co_err);
2188 kfree_skb(tunnel_buf);
2193 * Bundler functionality:
2195 void tipc_link_recv_bundle(struct sk_buff *buf)
2197 u32 msgcount = msg_msgcnt(buf_msg(buf));
2198 u32 pos = INT_H_SIZE;
2199 struct sk_buff *obuf;
2201 while (msgcount--) {
2202 obuf = buf_extract(buf, pos);
2204 pr_warn("Link unable to unbundle message(s)\n");
2207 pos += align(msg_size(buf_msg(obuf)));
2208 tipc_net_route_msg(obuf);
2214 * Fragmentation/defragmentation:
2218 * link_send_long_buf: Entry for buffers needing fragmentation.
2219 * The buffer is complete, inclusive total message length.
2220 * Returns user data length.
2222 static int link_send_long_buf(struct tipc_link *l_ptr, struct sk_buff *buf)
2224 struct sk_buff *buf_chain = NULL;
2225 struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2226 struct tipc_msg *inmsg = buf_msg(buf);
2227 struct tipc_msg fragm_hdr;
2228 u32 insize = msg_size(inmsg);
2229 u32 dsz = msg_data_sz(inmsg);
2230 unchar *crs = buf->data;
2232 u32 pack_sz = l_ptr->max_pkt;
2233 u32 fragm_sz = pack_sz - INT_H_SIZE;
2237 if (msg_short(inmsg))
2238 destaddr = l_ptr->addr;
2240 destaddr = msg_destnode(inmsg);
2242 /* Prepare reusable fragment header: */
2243 tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2244 INT_H_SIZE, destaddr);
2246 /* Chop up message: */
2248 struct sk_buff *fragm;
2250 if (rest <= fragm_sz) {
2252 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2254 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2255 if (fragm == NULL) {
2257 kfree_skb_list(buf_chain);
2260 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2262 msg_set_fragm_no(&fragm_hdr, fragm_no);
2263 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2264 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2266 buf_chain_tail->next = fragm;
2267 buf_chain_tail = fragm;
2271 msg_set_type(&fragm_hdr, FRAGMENT);
2275 /* Append chain of fragments to send queue & send them */
2276 l_ptr->long_msg_seq_no++;
2277 link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2278 l_ptr->stats.sent_fragments += fragm_no;
2279 l_ptr->stats.sent_fragmented++;
2280 tipc_link_push_queue(l_ptr);
2286 * tipc_link_recv_fragment(): Called with node lock on. Returns
2287 * the reassembled buffer if message is complete.
2289 int tipc_link_recv_fragment(struct sk_buff **head, struct sk_buff **tail,
2290 struct sk_buff **fbuf)
2292 struct sk_buff *frag = *fbuf;
2293 struct tipc_msg *msg = buf_msg(frag);
2294 u32 fragid = msg_type(msg);
2298 skb_pull(frag, msg_hdr_sz(msg));
2299 if (fragid == FIRST_FRAGMENT) {
2300 if (*head || skb_unclone(frag, GFP_ATOMIC))
2303 skb_frag_list_init(*head);
2306 skb_try_coalesce(*head, frag, &headstolen, &delta)) {
2307 kfree_skb_partial(frag, headstolen);
2311 if (!skb_has_frag_list(*head))
2312 skb_shinfo(*head)->frag_list = frag;
2314 (*tail)->next = frag;
2316 (*head)->truesize += frag->truesize;
2318 if (fragid == LAST_FRAGMENT) {
2320 *tail = *head = NULL;
2321 return LINK_REASM_COMPLETE;
2325 pr_warn_ratelimited("Link unable to reassemble fragmented message\n");
2327 return LINK_REASM_ERROR;
2330 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
2332 if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2335 l_ptr->tolerance = tolerance;
2336 l_ptr->continuity_interval =
2337 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2338 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2341 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
2343 /* Data messages from this node, inclusive FIRST_FRAGM */
2344 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2345 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2346 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2347 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2348 /* Transiting data messages,inclusive FIRST_FRAGM */
2349 l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2350 l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2351 l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2352 l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2353 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2354 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2355 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2356 /* FRAGMENT and LAST_FRAGMENT packets */
2357 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2361 * link_find_link - locate link by name
2362 * @name: ptr to link name string
2363 * @node: ptr to area to be filled with ptr to associated node
2365 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2366 * this also prevents link deletion.
2368 * Returns pointer to link (or 0 if invalid link name).
2370 static struct tipc_link *link_find_link(const char *name,
2371 struct tipc_node **node)
2373 struct tipc_link *l_ptr;
2374 struct tipc_node *n_ptr;
2377 list_for_each_entry(n_ptr, &tipc_node_list, list) {
2378 for (i = 0; i < MAX_BEARERS; i++) {
2379 l_ptr = n_ptr->links[i];
2380 if (l_ptr && !strcmp(l_ptr->name, name))
2392 * link_value_is_valid -- validate proposed link tolerance/priority/window
2394 * @cmd: value type (TIPC_CMD_SET_LINK_*)
2395 * @new_value: the new value
2397 * Returns 1 if value is within range, 0 if not.
2399 static int link_value_is_valid(u16 cmd, u32 new_value)
2402 case TIPC_CMD_SET_LINK_TOL:
2403 return (new_value >= TIPC_MIN_LINK_TOL) &&
2404 (new_value <= TIPC_MAX_LINK_TOL);
2405 case TIPC_CMD_SET_LINK_PRI:
2406 return (new_value <= TIPC_MAX_LINK_PRI);
2407 case TIPC_CMD_SET_LINK_WINDOW:
2408 return (new_value >= TIPC_MIN_LINK_WIN) &&
2409 (new_value <= TIPC_MAX_LINK_WIN);
2415 * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2416 * @name: ptr to link, bearer, or media name
2417 * @new_value: new value of link, bearer, or media setting
2418 * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2420 * Caller must hold 'tipc_net_lock' to ensure link/bearer/media is not deleted.
2422 * Returns 0 if value updated and negative value on error.
2424 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2426 struct tipc_node *node;
2427 struct tipc_link *l_ptr;
2428 struct tipc_bearer *b_ptr;
2429 struct tipc_media *m_ptr;
2432 l_ptr = link_find_link(name, &node);
2435 * acquire node lock for tipc_link_send_proto_msg().
2436 * see "TIPC locking policy" in net.c.
2438 tipc_node_lock(node);
2440 case TIPC_CMD_SET_LINK_TOL:
2441 link_set_supervision_props(l_ptr, new_value);
2442 tipc_link_send_proto_msg(l_ptr,
2443 STATE_MSG, 0, 0, new_value, 0, 0);
2445 case TIPC_CMD_SET_LINK_PRI:
2446 l_ptr->priority = new_value;
2447 tipc_link_send_proto_msg(l_ptr,
2448 STATE_MSG, 0, 0, 0, new_value, 0);
2450 case TIPC_CMD_SET_LINK_WINDOW:
2451 tipc_link_set_queue_limits(l_ptr, new_value);
2457 tipc_node_unlock(node);
2461 b_ptr = tipc_bearer_find(name);
2464 case TIPC_CMD_SET_LINK_TOL:
2465 b_ptr->tolerance = new_value;
2467 case TIPC_CMD_SET_LINK_PRI:
2468 b_ptr->priority = new_value;
2470 case TIPC_CMD_SET_LINK_WINDOW:
2471 b_ptr->window = new_value;
2480 m_ptr = tipc_media_find(name);
2484 case TIPC_CMD_SET_LINK_TOL:
2485 m_ptr->tolerance = new_value;
2487 case TIPC_CMD_SET_LINK_PRI:
2488 m_ptr->priority = new_value;
2490 case TIPC_CMD_SET_LINK_WINDOW:
2491 m_ptr->window = new_value;
2500 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2503 struct tipc_link_config *args;
2507 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2508 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2510 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2511 new_value = ntohl(args->value);
2513 if (!link_value_is_valid(cmd, new_value))
2514 return tipc_cfg_reply_error_string(
2515 "cannot change, value invalid");
2517 if (!strcmp(args->name, tipc_bclink_name)) {
2518 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2519 (tipc_bclink_set_queue_limits(new_value) == 0))
2520 return tipc_cfg_reply_none();
2521 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2522 " (cannot change setting on broadcast link)");
2525 read_lock_bh(&tipc_net_lock);
2526 res = link_cmd_set_value(args->name, new_value, cmd);
2527 read_unlock_bh(&tipc_net_lock);
2529 return tipc_cfg_reply_error_string("cannot change link setting");
2531 return tipc_cfg_reply_none();
2535 * link_reset_statistics - reset link statistics
2536 * @l_ptr: pointer to link
2538 static void link_reset_statistics(struct tipc_link *l_ptr)
2540 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2541 l_ptr->stats.sent_info = l_ptr->next_out_no;
2542 l_ptr->stats.recv_info = l_ptr->next_in_no;
2545 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2548 struct tipc_link *l_ptr;
2549 struct tipc_node *node;
2551 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2552 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2554 link_name = (char *)TLV_DATA(req_tlv_area);
2555 if (!strcmp(link_name, tipc_bclink_name)) {
2556 if (tipc_bclink_reset_stats())
2557 return tipc_cfg_reply_error_string("link not found");
2558 return tipc_cfg_reply_none();
2561 read_lock_bh(&tipc_net_lock);
2562 l_ptr = link_find_link(link_name, &node);
2564 read_unlock_bh(&tipc_net_lock);
2565 return tipc_cfg_reply_error_string("link not found");
2568 tipc_node_lock(node);
2569 link_reset_statistics(l_ptr);
2570 tipc_node_unlock(node);
2571 read_unlock_bh(&tipc_net_lock);
2572 return tipc_cfg_reply_none();
2576 * percent - convert count to a percentage of total (rounding up or down)
2578 static u32 percent(u32 count, u32 total)
2580 return (count * 100 + (total / 2)) / total;
2584 * tipc_link_stats - print link statistics
2586 * @buf: print buffer area
2587 * @buf_size: size of print buffer area
2589 * Returns length of print buffer data string (or 0 if error)
2591 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2593 struct tipc_link *l;
2594 struct tipc_stats *s;
2595 struct tipc_node *node;
2597 u32 profile_total = 0;
2600 if (!strcmp(name, tipc_bclink_name))
2601 return tipc_bclink_stats(buf, buf_size);
2603 read_lock_bh(&tipc_net_lock);
2604 l = link_find_link(name, &node);
2606 read_unlock_bh(&tipc_net_lock);
2609 tipc_node_lock(node);
2612 if (tipc_link_is_active(l))
2614 else if (tipc_link_is_up(l))
2619 ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2620 " %s MTU:%u Priority:%u Tolerance:%u ms"
2621 " Window:%u packets\n",
2622 l->name, status, l->max_pkt, l->priority,
2623 l->tolerance, l->queue_limit[0]);
2625 ret += tipc_snprintf(buf + ret, buf_size - ret,
2626 " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2627 l->next_in_no - s->recv_info, s->recv_fragments,
2628 s->recv_fragmented, s->recv_bundles,
2631 ret += tipc_snprintf(buf + ret, buf_size - ret,
2632 " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2633 l->next_out_no - s->sent_info, s->sent_fragments,
2634 s->sent_fragmented, s->sent_bundles,
2637 profile_total = s->msg_length_counts;
2641 ret += tipc_snprintf(buf + ret, buf_size - ret,
2642 " TX profile sample:%u packets average:%u octets\n"
2643 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2644 "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2645 s->msg_length_counts,
2646 s->msg_lengths_total / profile_total,
2647 percent(s->msg_length_profile[0], profile_total),
2648 percent(s->msg_length_profile[1], profile_total),
2649 percent(s->msg_length_profile[2], profile_total),
2650 percent(s->msg_length_profile[3], profile_total),
2651 percent(s->msg_length_profile[4], profile_total),
2652 percent(s->msg_length_profile[5], profile_total),
2653 percent(s->msg_length_profile[6], profile_total));
2655 ret += tipc_snprintf(buf + ret, buf_size - ret,
2656 " RX states:%u probes:%u naks:%u defs:%u"
2657 " dups:%u\n", s->recv_states, s->recv_probes,
2658 s->recv_nacks, s->deferred_recv, s->duplicates);
2660 ret += tipc_snprintf(buf + ret, buf_size - ret,
2661 " TX states:%u probes:%u naks:%u acks:%u"
2662 " dups:%u\n", s->sent_states, s->sent_probes,
2663 s->sent_nacks, s->sent_acks, s->retransmitted);
2665 ret += tipc_snprintf(buf + ret, buf_size - ret,
2666 " Congestion link:%u Send queue"
2667 " max:%u avg:%u\n", s->link_congs,
2668 s->max_queue_sz, s->queue_sz_counts ?
2669 (s->accu_queue_sz / s->queue_sz_counts) : 0);
2671 tipc_node_unlock(node);
2672 read_unlock_bh(&tipc_net_lock);
2676 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2678 struct sk_buff *buf;
2679 struct tlv_desc *rep_tlv;
2684 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2685 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2687 buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2691 rep_tlv = (struct tlv_desc *)buf->data;
2692 pb = TLV_DATA(rep_tlv);
2693 pb_len = ULTRA_STRING_MAX_LEN;
2694 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2698 return tipc_cfg_reply_error_string("link not found");
2700 str_len += 1; /* for "\0" */
2701 skb_put(buf, TLV_SPACE(str_len));
2702 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2708 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2709 * @dest: network address of destination node
2710 * @selector: used to select from set of active links
2712 * If no active link can be found, uses default maximum packet size.
2714 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2716 struct tipc_node *n_ptr;
2717 struct tipc_link *l_ptr;
2718 u32 res = MAX_PKT_DEFAULT;
2720 if (dest == tipc_own_addr)
2721 return MAX_MSG_SIZE;
2723 read_lock_bh(&tipc_net_lock);
2724 n_ptr = tipc_node_find(dest);
2726 tipc_node_lock(n_ptr);
2727 l_ptr = n_ptr->active_links[selector & 1];
2729 res = l_ptr->max_pkt;
2730 tipc_node_unlock(n_ptr);
2732 read_unlock_bh(&tipc_net_lock);
2736 static void link_print(struct tipc_link *l_ptr, const char *str)
2738 pr_info("%s Link %x<%s>:", str, l_ptr->addr, l_ptr->b_ptr->name);
2740 if (link_working_unknown(l_ptr))
2742 else if (link_reset_reset(l_ptr))
2744 else if (link_reset_unknown(l_ptr))
2746 else if (link_working_working(l_ptr))