writeback: fix time ordering of the per superblock dirty inode lists 6
[linux-2.6/zen-sources.git] / net / tipc / link.c
blob1b17fecee74784cd2a2f701a374eb0769607fb29
1 /*
2 * net/tipc/link.c: TIPC link code
4 * Copyright (c) 1996-2007, Ericsson AB
5 * Copyright (c) 2004-2007, Wind River Systems
6 * All rights reserved.
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.
37 #include "core.h"
38 #include "dbg.h"
39 #include "link.h"
40 #include "net.h"
41 #include "node.h"
42 #include "port.h"
43 #include "addr.h"
44 #include "node_subscr.h"
45 #include "name_distr.h"
46 #include "bearer.h"
47 #include "name_table.h"
48 #include "discover.h"
49 #include "config.h"
50 #include "bcast.h"
54 * Limit for deferred reception queue:
57 #define DEF_QUEUE_LIMIT 256u
60 * Link state events:
63 #define STARTING_EVT 856384768 /* link processing trigger */
64 #define TRAFFIC_MSG_EVT 560815u /* rx'd ??? */
65 #define TIMEOUT_EVT 560817u /* link timer expired */
68 * The following two 'message types' is really just implementation
69 * data conveniently stored in the message header.
70 * They must not be considered part of the protocol
72 #define OPEN_MSG 0
73 #define CLOSED_MSG 1
76 * State value stored in 'exp_msg_count'
79 #define START_CHANGEOVER 100000u
81 /**
82 * struct link_name - deconstructed link name
83 * @addr_local: network address of node at this end
84 * @if_local: name of interface at this end
85 * @addr_peer: network address of node at far end
86 * @if_peer: name of interface at far end
89 struct link_name {
90 u32 addr_local;
91 char if_local[TIPC_MAX_IF_NAME];
92 u32 addr_peer;
93 char if_peer[TIPC_MAX_IF_NAME];
96 #if 0
98 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
101 * struct link_event - link up/down event notification
104 struct link_event {
105 u32 addr;
106 int up;
107 void (*fcn)(u32, char *, int);
108 char name[TIPC_MAX_LINK_NAME];
111 #endif
113 static void link_handle_out_of_seq_msg(struct link *l_ptr,
114 struct sk_buff *buf);
115 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf);
116 static int link_recv_changeover_msg(struct link **l_ptr, struct sk_buff **buf);
117 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance);
118 static int link_send_sections_long(struct port *sender,
119 struct iovec const *msg_sect,
120 u32 num_sect, u32 destnode);
121 static void link_check_defragm_bufs(struct link *l_ptr);
122 static void link_state_event(struct link *l_ptr, u32 event);
123 static void link_reset_statistics(struct link *l_ptr);
124 static void link_print(struct link *l_ptr, struct print_buf *buf,
125 const char *str);
128 * Debugging code used by link routines only
130 * When debugging link problems on a system that has multiple links,
131 * the standard TIPC debugging routines may not be useful since they
132 * allow the output from multiple links to be intermixed. For this reason
133 * routines of the form "dbg_link_XXX()" have been created that will capture
134 * debug info into a link's personal print buffer, which can then be dumped
135 * into the TIPC system log (TIPC_LOG) upon request.
137 * To enable per-link debugging, use LINK_LOG_BUF_SIZE to specify the size
138 * of the print buffer used by each link. If LINK_LOG_BUF_SIZE is set to 0,
139 * the dbg_link_XXX() routines simply send their output to the standard
140 * debug print buffer (DBG_OUTPUT), if it has been defined; this can be useful
141 * when there is only a single link in the system being debugged.
143 * Notes:
144 * - When enabled, LINK_LOG_BUF_SIZE should be set to at least TIPC_PB_MIN_SIZE
145 * - "l_ptr" must be valid when using dbg_link_XXX() macros
148 #define LINK_LOG_BUF_SIZE 0
150 #define dbg_link(fmt, arg...) do {if (LINK_LOG_BUF_SIZE) tipc_printf(&l_ptr->print_buf, fmt, ## arg); } while(0)
151 #define dbg_link_msg(msg, txt) do {if (LINK_LOG_BUF_SIZE) tipc_msg_print(&l_ptr->print_buf, msg, txt); } while(0)
152 #define dbg_link_state(txt) do {if (LINK_LOG_BUF_SIZE) link_print(l_ptr, &l_ptr->print_buf, txt); } while(0)
153 #define dbg_link_dump() do { \
154 if (LINK_LOG_BUF_SIZE) { \
155 tipc_printf(LOG, "\n\nDumping link <%s>:\n", l_ptr->name); \
156 tipc_printbuf_move(LOG, &l_ptr->print_buf); \
158 } while (0)
160 static void dbg_print_link(struct link *l_ptr, const char *str)
162 if (DBG_OUTPUT != TIPC_NULL)
163 link_print(l_ptr, DBG_OUTPUT, str);
166 static void dbg_print_buf_chain(struct sk_buff *root_buf)
168 if (DBG_OUTPUT != TIPC_NULL) {
169 struct sk_buff *buf = root_buf;
171 while (buf) {
172 msg_dbg(buf_msg(buf), "In chain: ");
173 buf = buf->next;
179 * Simple link routines
182 static unsigned int align(unsigned int i)
184 return (i + 3) & ~3u;
187 static int link_working_working(struct link *l_ptr)
189 return (l_ptr->state == WORKING_WORKING);
192 static int link_working_unknown(struct link *l_ptr)
194 return (l_ptr->state == WORKING_UNKNOWN);
197 static int link_reset_unknown(struct link *l_ptr)
199 return (l_ptr->state == RESET_UNKNOWN);
202 static int link_reset_reset(struct link *l_ptr)
204 return (l_ptr->state == RESET_RESET);
207 static int link_blocked(struct link *l_ptr)
209 return (l_ptr->exp_msg_count || l_ptr->blocked);
212 static int link_congested(struct link *l_ptr)
214 return (l_ptr->out_queue_size >= l_ptr->queue_limit[0]);
217 static u32 link_max_pkt(struct link *l_ptr)
219 return l_ptr->max_pkt;
222 static void link_init_max_pkt(struct link *l_ptr)
224 u32 max_pkt;
226 max_pkt = (l_ptr->b_ptr->publ.mtu & ~3);
227 if (max_pkt > MAX_MSG_SIZE)
228 max_pkt = MAX_MSG_SIZE;
230 l_ptr->max_pkt_target = max_pkt;
231 if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
232 l_ptr->max_pkt = l_ptr->max_pkt_target;
233 else
234 l_ptr->max_pkt = MAX_PKT_DEFAULT;
236 l_ptr->max_pkt_probes = 0;
239 static u32 link_next_sent(struct link *l_ptr)
241 if (l_ptr->next_out)
242 return msg_seqno(buf_msg(l_ptr->next_out));
243 return mod(l_ptr->next_out_no);
246 static u32 link_last_sent(struct link *l_ptr)
248 return mod(link_next_sent(l_ptr) - 1);
252 * Simple non-static link routines (i.e. referenced outside this file)
255 int tipc_link_is_up(struct link *l_ptr)
257 if (!l_ptr)
258 return 0;
259 return (link_working_working(l_ptr) || link_working_unknown(l_ptr));
262 int tipc_link_is_active(struct link *l_ptr)
264 return ((l_ptr->owner->active_links[0] == l_ptr) ||
265 (l_ptr->owner->active_links[1] == l_ptr));
269 * link_name_validate - validate & (optionally) deconstruct link name
270 * @name - ptr to link name string
271 * @name_parts - ptr to area for link name components (or NULL if not needed)
273 * Returns 1 if link name is valid, otherwise 0.
276 static int link_name_validate(const char *name, struct link_name *name_parts)
278 char name_copy[TIPC_MAX_LINK_NAME];
279 char *addr_local;
280 char *if_local;
281 char *addr_peer;
282 char *if_peer;
283 char dummy;
284 u32 z_local, c_local, n_local;
285 u32 z_peer, c_peer, n_peer;
286 u32 if_local_len;
287 u32 if_peer_len;
289 /* copy link name & ensure length is OK */
291 name_copy[TIPC_MAX_LINK_NAME - 1] = 0;
292 /* need above in case non-Posix strncpy() doesn't pad with nulls */
293 strncpy(name_copy, name, TIPC_MAX_LINK_NAME);
294 if (name_copy[TIPC_MAX_LINK_NAME - 1] != 0)
295 return 0;
297 /* ensure all component parts of link name are present */
299 addr_local = name_copy;
300 if ((if_local = strchr(addr_local, ':')) == NULL)
301 return 0;
302 *(if_local++) = 0;
303 if ((addr_peer = strchr(if_local, '-')) == NULL)
304 return 0;
305 *(addr_peer++) = 0;
306 if_local_len = addr_peer - if_local;
307 if ((if_peer = strchr(addr_peer, ':')) == NULL)
308 return 0;
309 *(if_peer++) = 0;
310 if_peer_len = strlen(if_peer) + 1;
312 /* validate component parts of link name */
314 if ((sscanf(addr_local, "%u.%u.%u%c",
315 &z_local, &c_local, &n_local, &dummy) != 3) ||
316 (sscanf(addr_peer, "%u.%u.%u%c",
317 &z_peer, &c_peer, &n_peer, &dummy) != 3) ||
318 (z_local > 255) || (c_local > 4095) || (n_local > 4095) ||
319 (z_peer > 255) || (c_peer > 4095) || (n_peer > 4095) ||
320 (if_local_len <= 1) || (if_local_len > TIPC_MAX_IF_NAME) ||
321 (if_peer_len <= 1) || (if_peer_len > TIPC_MAX_IF_NAME) ||
322 (strspn(if_local, tipc_alphabet) != (if_local_len - 1)) ||
323 (strspn(if_peer, tipc_alphabet) != (if_peer_len - 1)))
324 return 0;
326 /* return link name components, if necessary */
328 if (name_parts) {
329 name_parts->addr_local = tipc_addr(z_local, c_local, n_local);
330 strcpy(name_parts->if_local, if_local);
331 name_parts->addr_peer = tipc_addr(z_peer, c_peer, n_peer);
332 strcpy(name_parts->if_peer, if_peer);
334 return 1;
338 * link_timeout - handle expiration of link timer
339 * @l_ptr: pointer to link
341 * This routine must not grab "tipc_net_lock" to avoid a potential deadlock conflict
342 * with tipc_link_delete(). (There is no risk that the node will be deleted by
343 * another thread because tipc_link_delete() always cancels the link timer before
344 * tipc_node_delete() is called.)
347 static void link_timeout(struct link *l_ptr)
349 tipc_node_lock(l_ptr->owner);
351 /* update counters used in statistical profiling of send traffic */
353 l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
354 l_ptr->stats.queue_sz_counts++;
356 if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
357 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
359 if (l_ptr->first_out) {
360 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
361 u32 length = msg_size(msg);
363 if ((msg_user(msg) == MSG_FRAGMENTER)
364 && (msg_type(msg) == FIRST_FRAGMENT)) {
365 length = msg_size(msg_get_wrapped(msg));
367 if (length) {
368 l_ptr->stats.msg_lengths_total += length;
369 l_ptr->stats.msg_length_counts++;
370 if (length <= 64)
371 l_ptr->stats.msg_length_profile[0]++;
372 else if (length <= 256)
373 l_ptr->stats.msg_length_profile[1]++;
374 else if (length <= 1024)
375 l_ptr->stats.msg_length_profile[2]++;
376 else if (length <= 4096)
377 l_ptr->stats.msg_length_profile[3]++;
378 else if (length <= 16384)
379 l_ptr->stats.msg_length_profile[4]++;
380 else if (length <= 32768)
381 l_ptr->stats.msg_length_profile[5]++;
382 else
383 l_ptr->stats.msg_length_profile[6]++;
387 /* do all other link processing performed on a periodic basis */
389 link_check_defragm_bufs(l_ptr);
391 link_state_event(l_ptr, TIMEOUT_EVT);
393 if (l_ptr->next_out)
394 tipc_link_push_queue(l_ptr);
396 tipc_node_unlock(l_ptr->owner);
399 static void link_set_timer(struct link *l_ptr, u32 time)
401 k_start_timer(&l_ptr->timer, time);
405 * tipc_link_create - create a new link
406 * @b_ptr: pointer to associated bearer
407 * @peer: network address of node at other end of link
408 * @media_addr: media address to use when sending messages over link
410 * Returns pointer to link.
413 struct link *tipc_link_create(struct bearer *b_ptr, const u32 peer,
414 const struct tipc_media_addr *media_addr)
416 struct link *l_ptr;
417 struct tipc_msg *msg;
418 char *if_name;
420 l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
421 if (!l_ptr) {
422 warn("Link creation failed, no memory\n");
423 return NULL;
426 if (LINK_LOG_BUF_SIZE) {
427 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
429 if (!pb) {
430 kfree(l_ptr);
431 warn("Link creation failed, no memory for print buffer\n");
432 return NULL;
434 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
437 l_ptr->addr = peer;
438 if_name = strchr(b_ptr->publ.name, ':') + 1;
439 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
440 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
441 tipc_node(tipc_own_addr),
442 if_name,
443 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
444 /* note: peer i/f is appended to link name by reset/activate */
445 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
446 l_ptr->checkpoint = 1;
447 l_ptr->b_ptr = b_ptr;
448 link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
449 l_ptr->state = RESET_UNKNOWN;
451 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
452 msg = l_ptr->pmsg;
453 msg_init(msg, LINK_PROTOCOL, RESET_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
454 msg_set_size(msg, sizeof(l_ptr->proto_msg));
455 msg_set_session(msg, tipc_random);
456 msg_set_bearer_id(msg, b_ptr->identity);
457 strcpy((char *)msg_data(msg), if_name);
459 l_ptr->priority = b_ptr->priority;
460 tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
462 link_init_max_pkt(l_ptr);
464 l_ptr->next_out_no = 1;
465 INIT_LIST_HEAD(&l_ptr->waiting_ports);
467 link_reset_statistics(l_ptr);
469 l_ptr->owner = tipc_node_attach_link(l_ptr);
470 if (!l_ptr->owner) {
471 if (LINK_LOG_BUF_SIZE)
472 kfree(l_ptr->print_buf.buf);
473 kfree(l_ptr);
474 return NULL;
477 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
478 list_add_tail(&l_ptr->link_list, &b_ptr->links);
479 tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
481 dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
482 l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
484 return l_ptr;
488 * tipc_link_delete - delete a link
489 * @l_ptr: pointer to link
491 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
492 * This routine must not grab the node lock until after link timer cancellation
493 * to avoid a potential deadlock situation.
496 void tipc_link_delete(struct link *l_ptr)
498 if (!l_ptr) {
499 err("Attempt to delete non-existent link\n");
500 return;
503 dbg("tipc_link_delete()\n");
505 k_cancel_timer(&l_ptr->timer);
507 tipc_node_lock(l_ptr->owner);
508 tipc_link_reset(l_ptr);
509 tipc_node_detach_link(l_ptr->owner, l_ptr);
510 tipc_link_stop(l_ptr);
511 list_del_init(&l_ptr->link_list);
512 if (LINK_LOG_BUF_SIZE)
513 kfree(l_ptr->print_buf.buf);
514 tipc_node_unlock(l_ptr->owner);
515 k_term_timer(&l_ptr->timer);
516 kfree(l_ptr);
519 void tipc_link_start(struct link *l_ptr)
521 dbg("tipc_link_start %x\n", l_ptr);
522 link_state_event(l_ptr, STARTING_EVT);
526 * link_schedule_port - schedule port for deferred sending
527 * @l_ptr: pointer to link
528 * @origport: reference to sending port
529 * @sz: amount of data to be sent
531 * Schedules port for renewed sending of messages after link congestion
532 * has abated.
535 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
537 struct port *p_ptr;
539 spin_lock_bh(&tipc_port_list_lock);
540 p_ptr = tipc_port_lock(origport);
541 if (p_ptr) {
542 if (!p_ptr->wakeup)
543 goto exit;
544 if (!list_empty(&p_ptr->wait_list))
545 goto exit;
546 p_ptr->congested_link = l_ptr;
547 p_ptr->publ.congested = 1;
548 p_ptr->waiting_pkts = 1 + ((sz - 1) / link_max_pkt(l_ptr));
549 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
550 l_ptr->stats.link_congs++;
551 exit:
552 tipc_port_unlock(p_ptr);
554 spin_unlock_bh(&tipc_port_list_lock);
555 return -ELINKCONG;
558 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
560 struct port *p_ptr;
561 struct port *temp_p_ptr;
562 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
564 if (all)
565 win = 100000;
566 if (win <= 0)
567 return;
568 if (!spin_trylock_bh(&tipc_port_list_lock))
569 return;
570 if (link_congested(l_ptr))
571 goto exit;
572 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
573 wait_list) {
574 if (win <= 0)
575 break;
576 list_del_init(&p_ptr->wait_list);
577 p_ptr->congested_link = NULL;
578 spin_lock_bh(p_ptr->publ.lock);
579 p_ptr->publ.congested = 0;
580 p_ptr->wakeup(&p_ptr->publ);
581 win -= p_ptr->waiting_pkts;
582 spin_unlock_bh(p_ptr->publ.lock);
585 exit:
586 spin_unlock_bh(&tipc_port_list_lock);
590 * link_release_outqueue - purge link's outbound message queue
591 * @l_ptr: pointer to link
594 static void link_release_outqueue(struct link *l_ptr)
596 struct sk_buff *buf = l_ptr->first_out;
597 struct sk_buff *next;
599 while (buf) {
600 next = buf->next;
601 buf_discard(buf);
602 buf = next;
604 l_ptr->first_out = NULL;
605 l_ptr->out_queue_size = 0;
609 * tipc_link_reset_fragments - purge link's inbound message fragments queue
610 * @l_ptr: pointer to link
613 void tipc_link_reset_fragments(struct link *l_ptr)
615 struct sk_buff *buf = l_ptr->defragm_buf;
616 struct sk_buff *next;
618 while (buf) {
619 next = buf->next;
620 buf_discard(buf);
621 buf = next;
623 l_ptr->defragm_buf = NULL;
627 * tipc_link_stop - purge all inbound and outbound messages associated with link
628 * @l_ptr: pointer to link
631 void tipc_link_stop(struct link *l_ptr)
633 struct sk_buff *buf;
634 struct sk_buff *next;
636 buf = l_ptr->oldest_deferred_in;
637 while (buf) {
638 next = buf->next;
639 buf_discard(buf);
640 buf = next;
643 buf = l_ptr->first_out;
644 while (buf) {
645 next = buf->next;
646 buf_discard(buf);
647 buf = next;
650 tipc_link_reset_fragments(l_ptr);
652 buf_discard(l_ptr->proto_msg_queue);
653 l_ptr->proto_msg_queue = NULL;
656 #if 0
658 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
660 static void link_recv_event(struct link_event *ev)
662 ev->fcn(ev->addr, ev->name, ev->up);
663 kfree(ev);
666 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
667 struct link *l_ptr, int up)
669 struct link_event *ev;
671 ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
672 if (!ev) {
673 warn("Link event allocation failure\n");
674 return;
676 ev->addr = l_ptr->addr;
677 ev->up = up;
678 ev->fcn = fcn;
679 memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
680 tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
683 #else
685 #define link_send_event(fcn, l_ptr, up) do { } while (0)
687 #endif
689 void tipc_link_reset(struct link *l_ptr)
691 struct sk_buff *buf;
692 u32 prev_state = l_ptr->state;
693 u32 checkpoint = l_ptr->next_in_no;
694 int was_active_link = tipc_link_is_active(l_ptr);
696 msg_set_session(l_ptr->pmsg, msg_session(l_ptr->pmsg) + 1);
698 /* Link is down, accept any session: */
699 l_ptr->peer_session = 0;
701 /* Prepare for max packet size negotiation */
702 link_init_max_pkt(l_ptr);
704 l_ptr->state = RESET_UNKNOWN;
705 dbg_link_state("Resetting Link\n");
707 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
708 return;
710 tipc_node_link_down(l_ptr->owner, l_ptr);
711 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
712 #if 0
713 tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
714 dbg_link_dump();
715 #endif
716 if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
717 l_ptr->owner->permit_changeover) {
718 l_ptr->reset_checkpoint = checkpoint;
719 l_ptr->exp_msg_count = START_CHANGEOVER;
722 /* Clean up all queues: */
724 link_release_outqueue(l_ptr);
725 buf_discard(l_ptr->proto_msg_queue);
726 l_ptr->proto_msg_queue = NULL;
727 buf = l_ptr->oldest_deferred_in;
728 while (buf) {
729 struct sk_buff *next = buf->next;
730 buf_discard(buf);
731 buf = next;
733 if (!list_empty(&l_ptr->waiting_ports))
734 tipc_link_wakeup_ports(l_ptr, 1);
736 l_ptr->retransm_queue_head = 0;
737 l_ptr->retransm_queue_size = 0;
738 l_ptr->last_out = NULL;
739 l_ptr->first_out = NULL;
740 l_ptr->next_out = NULL;
741 l_ptr->unacked_window = 0;
742 l_ptr->checkpoint = 1;
743 l_ptr->next_out_no = 1;
744 l_ptr->deferred_inqueue_sz = 0;
745 l_ptr->oldest_deferred_in = NULL;
746 l_ptr->newest_deferred_in = NULL;
747 l_ptr->fsm_msg_cnt = 0;
748 l_ptr->stale_count = 0;
749 link_reset_statistics(l_ptr);
751 link_send_event(tipc_cfg_link_event, l_ptr, 0);
752 if (!in_own_cluster(l_ptr->addr))
753 link_send_event(tipc_disc_link_event, l_ptr, 0);
757 static void link_activate(struct link *l_ptr)
759 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
760 tipc_node_link_up(l_ptr->owner, l_ptr);
761 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
762 link_send_event(tipc_cfg_link_event, l_ptr, 1);
763 if (!in_own_cluster(l_ptr->addr))
764 link_send_event(tipc_disc_link_event, l_ptr, 1);
768 * link_state_event - link finite state machine
769 * @l_ptr: pointer to link
770 * @event: state machine event to process
773 static void link_state_event(struct link *l_ptr, unsigned event)
775 struct link *other;
776 u32 cont_intv = l_ptr->continuity_interval;
778 if (!l_ptr->started && (event != STARTING_EVT))
779 return; /* Not yet. */
781 if (link_blocked(l_ptr)) {
782 if (event == TIMEOUT_EVT) {
783 link_set_timer(l_ptr, cont_intv);
785 return; /* Changeover going on */
787 dbg_link("STATE_EV: <%s> ", l_ptr->name);
789 switch (l_ptr->state) {
790 case WORKING_WORKING:
791 dbg_link("WW/");
792 switch (event) {
793 case TRAFFIC_MSG_EVT:
794 dbg_link("TRF-");
795 /* fall through */
796 case ACTIVATE_MSG:
797 dbg_link("ACT\n");
798 break;
799 case TIMEOUT_EVT:
800 dbg_link("TIM ");
801 if (l_ptr->next_in_no != l_ptr->checkpoint) {
802 l_ptr->checkpoint = l_ptr->next_in_no;
803 if (tipc_bclink_acks_missing(l_ptr->owner)) {
804 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
805 0, 0, 0, 0, 0);
806 l_ptr->fsm_msg_cnt++;
807 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
808 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
809 1, 0, 0, 0, 0);
810 l_ptr->fsm_msg_cnt++;
812 link_set_timer(l_ptr, cont_intv);
813 break;
815 dbg_link(" -> WU\n");
816 l_ptr->state = WORKING_UNKNOWN;
817 l_ptr->fsm_msg_cnt = 0;
818 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
819 l_ptr->fsm_msg_cnt++;
820 link_set_timer(l_ptr, cont_intv / 4);
821 break;
822 case RESET_MSG:
823 dbg_link("RES -> RR\n");
824 info("Resetting link <%s>, requested by peer\n",
825 l_ptr->name);
826 tipc_link_reset(l_ptr);
827 l_ptr->state = RESET_RESET;
828 l_ptr->fsm_msg_cnt = 0;
829 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
830 l_ptr->fsm_msg_cnt++;
831 link_set_timer(l_ptr, cont_intv);
832 break;
833 default:
834 err("Unknown link event %u in WW state\n", event);
836 break;
837 case WORKING_UNKNOWN:
838 dbg_link("WU/");
839 switch (event) {
840 case TRAFFIC_MSG_EVT:
841 dbg_link("TRF-");
842 case ACTIVATE_MSG:
843 dbg_link("ACT -> WW\n");
844 l_ptr->state = WORKING_WORKING;
845 l_ptr->fsm_msg_cnt = 0;
846 link_set_timer(l_ptr, cont_intv);
847 break;
848 case RESET_MSG:
849 dbg_link("RES -> RR\n");
850 info("Resetting link <%s>, requested by peer "
851 "while probing\n", l_ptr->name);
852 tipc_link_reset(l_ptr);
853 l_ptr->state = RESET_RESET;
854 l_ptr->fsm_msg_cnt = 0;
855 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
856 l_ptr->fsm_msg_cnt++;
857 link_set_timer(l_ptr, cont_intv);
858 break;
859 case TIMEOUT_EVT:
860 dbg_link("TIM ");
861 if (l_ptr->next_in_no != l_ptr->checkpoint) {
862 dbg_link("-> WW \n");
863 l_ptr->state = WORKING_WORKING;
864 l_ptr->fsm_msg_cnt = 0;
865 l_ptr->checkpoint = l_ptr->next_in_no;
866 if (tipc_bclink_acks_missing(l_ptr->owner)) {
867 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
868 0, 0, 0, 0, 0);
869 l_ptr->fsm_msg_cnt++;
871 link_set_timer(l_ptr, cont_intv);
872 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
873 dbg_link("Probing %u/%u,timer = %u ms)\n",
874 l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
875 cont_intv / 4);
876 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
877 1, 0, 0, 0, 0);
878 l_ptr->fsm_msg_cnt++;
879 link_set_timer(l_ptr, cont_intv / 4);
880 } else { /* Link has failed */
881 dbg_link("-> RU (%u probes unanswered)\n",
882 l_ptr->fsm_msg_cnt);
883 warn("Resetting link <%s>, peer not responding\n",
884 l_ptr->name);
885 tipc_link_reset(l_ptr);
886 l_ptr->state = RESET_UNKNOWN;
887 l_ptr->fsm_msg_cnt = 0;
888 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
889 0, 0, 0, 0, 0);
890 l_ptr->fsm_msg_cnt++;
891 link_set_timer(l_ptr, cont_intv);
893 break;
894 default:
895 err("Unknown link event %u in WU state\n", event);
897 break;
898 case RESET_UNKNOWN:
899 dbg_link("RU/");
900 switch (event) {
901 case TRAFFIC_MSG_EVT:
902 dbg_link("TRF-\n");
903 break;
904 case ACTIVATE_MSG:
905 other = l_ptr->owner->active_links[0];
906 if (other && link_working_unknown(other)) {
907 dbg_link("ACT\n");
908 break;
910 dbg_link("ACT -> WW\n");
911 l_ptr->state = WORKING_WORKING;
912 l_ptr->fsm_msg_cnt = 0;
913 link_activate(l_ptr);
914 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
915 l_ptr->fsm_msg_cnt++;
916 link_set_timer(l_ptr, cont_intv);
917 break;
918 case RESET_MSG:
919 dbg_link("RES \n");
920 dbg_link(" -> RR\n");
921 l_ptr->state = RESET_RESET;
922 l_ptr->fsm_msg_cnt = 0;
923 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
924 l_ptr->fsm_msg_cnt++;
925 link_set_timer(l_ptr, cont_intv);
926 break;
927 case STARTING_EVT:
928 dbg_link("START-");
929 l_ptr->started = 1;
930 /* fall through */
931 case TIMEOUT_EVT:
932 dbg_link("TIM \n");
933 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
934 l_ptr->fsm_msg_cnt++;
935 link_set_timer(l_ptr, cont_intv);
936 break;
937 default:
938 err("Unknown link event %u in RU state\n", event);
940 break;
941 case RESET_RESET:
942 dbg_link("RR/ ");
943 switch (event) {
944 case TRAFFIC_MSG_EVT:
945 dbg_link("TRF-");
946 /* fall through */
947 case ACTIVATE_MSG:
948 other = l_ptr->owner->active_links[0];
949 if (other && link_working_unknown(other)) {
950 dbg_link("ACT\n");
951 break;
953 dbg_link("ACT -> WW\n");
954 l_ptr->state = WORKING_WORKING;
955 l_ptr->fsm_msg_cnt = 0;
956 link_activate(l_ptr);
957 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
958 l_ptr->fsm_msg_cnt++;
959 link_set_timer(l_ptr, cont_intv);
960 break;
961 case RESET_MSG:
962 dbg_link("RES\n");
963 break;
964 case TIMEOUT_EVT:
965 dbg_link("TIM\n");
966 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
967 l_ptr->fsm_msg_cnt++;
968 link_set_timer(l_ptr, cont_intv);
969 dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
970 break;
971 default:
972 err("Unknown link event %u in RR state\n", event);
974 break;
975 default:
976 err("Unknown link state %u/%u\n", l_ptr->state, event);
981 * link_bundle_buf(): Append contents of a buffer to
982 * the tail of an existing one.
985 static int link_bundle_buf(struct link *l_ptr,
986 struct sk_buff *bundler,
987 struct sk_buff *buf)
989 struct tipc_msg *bundler_msg = buf_msg(bundler);
990 struct tipc_msg *msg = buf_msg(buf);
991 u32 size = msg_size(msg);
992 u32 bundle_size = msg_size(bundler_msg);
993 u32 to_pos = align(bundle_size);
994 u32 pad = to_pos - bundle_size;
996 if (msg_user(bundler_msg) != MSG_BUNDLER)
997 return 0;
998 if (msg_type(bundler_msg) != OPEN_MSG)
999 return 0;
1000 if (skb_tailroom(bundler) < (pad + size))
1001 return 0;
1002 if (link_max_pkt(l_ptr) < (to_pos + size))
1003 return 0;
1005 skb_put(bundler, pad + size);
1006 skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
1007 msg_set_size(bundler_msg, to_pos + size);
1008 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
1009 dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
1010 msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
1011 msg_dbg(msg, "PACKD:");
1012 buf_discard(buf);
1013 l_ptr->stats.sent_bundled++;
1014 return 1;
1017 static void link_add_to_outqueue(struct link *l_ptr,
1018 struct sk_buff *buf,
1019 struct tipc_msg *msg)
1021 u32 ack = mod(l_ptr->next_in_no - 1);
1022 u32 seqno = mod(l_ptr->next_out_no++);
1024 msg_set_word(msg, 2, ((ack << 16) | seqno));
1025 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1026 buf->next = NULL;
1027 if (l_ptr->first_out) {
1028 l_ptr->last_out->next = buf;
1029 l_ptr->last_out = buf;
1030 } else
1031 l_ptr->first_out = l_ptr->last_out = buf;
1032 l_ptr->out_queue_size++;
1036 * tipc_link_send_buf() is the 'full path' for messages, called from
1037 * inside TIPC when the 'fast path' in tipc_send_buf
1038 * has failed, and from link_send()
1041 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1043 struct tipc_msg *msg = buf_msg(buf);
1044 u32 size = msg_size(msg);
1045 u32 dsz = msg_data_sz(msg);
1046 u32 queue_size = l_ptr->out_queue_size;
1047 u32 imp = msg_tot_importance(msg);
1048 u32 queue_limit = l_ptr->queue_limit[imp];
1049 u32 max_packet = link_max_pkt(l_ptr);
1051 msg_set_prevnode(msg, tipc_own_addr); /* If routed message */
1053 /* Match msg importance against queue limits: */
1055 if (unlikely(queue_size >= queue_limit)) {
1056 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1057 return link_schedule_port(l_ptr, msg_origport(msg),
1058 size);
1060 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1061 buf_discard(buf);
1062 if (imp > CONN_MANAGER) {
1063 warn("Resetting link <%s>, send queue full", l_ptr->name);
1064 tipc_link_reset(l_ptr);
1066 return dsz;
1069 /* Fragmentation needed ? */
1071 if (size > max_packet)
1072 return tipc_link_send_long_buf(l_ptr, buf);
1074 /* Packet can be queued or sent: */
1076 if (queue_size > l_ptr->stats.max_queue_sz)
1077 l_ptr->stats.max_queue_sz = queue_size;
1079 if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1080 !link_congested(l_ptr))) {
1081 link_add_to_outqueue(l_ptr, buf, msg);
1083 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1084 l_ptr->unacked_window = 0;
1085 } else {
1086 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1087 l_ptr->stats.bearer_congs++;
1088 l_ptr->next_out = buf;
1090 return dsz;
1092 /* Congestion: can message be bundled ?: */
1094 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1095 (msg_user(msg) != MSG_FRAGMENTER)) {
1097 /* Try adding message to an existing bundle */
1099 if (l_ptr->next_out &&
1100 link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1101 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1102 return dsz;
1105 /* Try creating a new bundle */
1107 if (size <= max_packet * 2 / 3) {
1108 struct sk_buff *bundler = buf_acquire(max_packet);
1109 struct tipc_msg bundler_hdr;
1111 if (bundler) {
1112 msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1113 TIPC_OK, INT_H_SIZE, l_ptr->addr);
1114 skb_copy_to_linear_data(bundler, &bundler_hdr,
1115 INT_H_SIZE);
1116 skb_trim(bundler, INT_H_SIZE);
1117 link_bundle_buf(l_ptr, bundler, buf);
1118 buf = bundler;
1119 msg = buf_msg(buf);
1120 l_ptr->stats.sent_bundles++;
1124 if (!l_ptr->next_out)
1125 l_ptr->next_out = buf;
1126 link_add_to_outqueue(l_ptr, buf, msg);
1127 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1128 return dsz;
1132 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1133 * not been selected yet, and the the owner node is not locked
1134 * Called by TIPC internal users, e.g. the name distributor
1137 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1139 struct link *l_ptr;
1140 struct node *n_ptr;
1141 int res = -ELINKCONG;
1143 read_lock_bh(&tipc_net_lock);
1144 n_ptr = tipc_node_select(dest, selector);
1145 if (n_ptr) {
1146 tipc_node_lock(n_ptr);
1147 l_ptr = n_ptr->active_links[selector & 1];
1148 if (l_ptr) {
1149 dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1150 res = tipc_link_send_buf(l_ptr, buf);
1151 } else {
1152 dbg("Attempt to send msg to unreachable node:\n");
1153 msg_dbg(buf_msg(buf),">>>");
1154 buf_discard(buf);
1156 tipc_node_unlock(n_ptr);
1157 } else {
1158 dbg("Attempt to send msg to unknown node:\n");
1159 msg_dbg(buf_msg(buf),">>>");
1160 buf_discard(buf);
1162 read_unlock_bh(&tipc_net_lock);
1163 return res;
1167 * link_send_buf_fast: Entry for data messages where the
1168 * destination link is known and the header is complete,
1169 * inclusive total message length. Very time critical.
1170 * Link is locked. Returns user data length.
1173 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1174 u32 *used_max_pkt)
1176 struct tipc_msg *msg = buf_msg(buf);
1177 int res = msg_data_sz(msg);
1179 if (likely(!link_congested(l_ptr))) {
1180 if (likely(msg_size(msg) <= link_max_pkt(l_ptr))) {
1181 if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1182 link_add_to_outqueue(l_ptr, buf, msg);
1183 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1184 &l_ptr->media_addr))) {
1185 l_ptr->unacked_window = 0;
1186 msg_dbg(msg,"SENT_FAST:");
1187 return res;
1189 dbg("failed sent fast...\n");
1190 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1191 l_ptr->stats.bearer_congs++;
1192 l_ptr->next_out = buf;
1193 return res;
1196 else
1197 *used_max_pkt = link_max_pkt(l_ptr);
1199 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
1203 * tipc_send_buf_fast: Entry for data messages where the
1204 * destination node is known and the header is complete,
1205 * inclusive total message length.
1206 * Returns user data length.
1208 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1210 struct link *l_ptr;
1211 struct node *n_ptr;
1212 int res;
1213 u32 selector = msg_origport(buf_msg(buf)) & 1;
1214 u32 dummy;
1216 if (destnode == tipc_own_addr)
1217 return tipc_port_recv_msg(buf);
1219 read_lock_bh(&tipc_net_lock);
1220 n_ptr = tipc_node_select(destnode, selector);
1221 if (likely(n_ptr)) {
1222 tipc_node_lock(n_ptr);
1223 l_ptr = n_ptr->active_links[selector];
1224 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1225 buf, l_ptr, destnode);
1226 if (likely(l_ptr)) {
1227 res = link_send_buf_fast(l_ptr, buf, &dummy);
1228 tipc_node_unlock(n_ptr);
1229 read_unlock_bh(&tipc_net_lock);
1230 return res;
1232 tipc_node_unlock(n_ptr);
1234 read_unlock_bh(&tipc_net_lock);
1235 res = msg_data_sz(buf_msg(buf));
1236 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1237 return res;
1242 * tipc_link_send_sections_fast: Entry for messages where the
1243 * destination processor is known and the header is complete,
1244 * except for total message length.
1245 * Returns user data length or errno.
1247 int tipc_link_send_sections_fast(struct port *sender,
1248 struct iovec const *msg_sect,
1249 const u32 num_sect,
1250 u32 destaddr)
1252 struct tipc_msg *hdr = &sender->publ.phdr;
1253 struct link *l_ptr;
1254 struct sk_buff *buf;
1255 struct node *node;
1256 int res;
1257 u32 selector = msg_origport(hdr) & 1;
1259 again:
1261 * Try building message using port's max_pkt hint.
1262 * (Must not hold any locks while building message.)
1265 res = msg_build(hdr, msg_sect, num_sect, sender->publ.max_pkt,
1266 !sender->user_port, &buf);
1268 read_lock_bh(&tipc_net_lock);
1269 node = tipc_node_select(destaddr, selector);
1270 if (likely(node)) {
1271 tipc_node_lock(node);
1272 l_ptr = node->active_links[selector];
1273 if (likely(l_ptr)) {
1274 if (likely(buf)) {
1275 res = link_send_buf_fast(l_ptr, buf,
1276 &sender->publ.max_pkt);
1277 if (unlikely(res < 0))
1278 buf_discard(buf);
1279 exit:
1280 tipc_node_unlock(node);
1281 read_unlock_bh(&tipc_net_lock);
1282 return res;
1285 /* Exit if build request was invalid */
1287 if (unlikely(res < 0))
1288 goto exit;
1290 /* Exit if link (or bearer) is congested */
1292 if (link_congested(l_ptr) ||
1293 !list_empty(&l_ptr->b_ptr->cong_links)) {
1294 res = link_schedule_port(l_ptr,
1295 sender->publ.ref, res);
1296 goto exit;
1300 * Message size exceeds max_pkt hint; update hint,
1301 * then re-try fast path or fragment the message
1304 sender->publ.max_pkt = link_max_pkt(l_ptr);
1305 tipc_node_unlock(node);
1306 read_unlock_bh(&tipc_net_lock);
1309 if ((msg_hdr_sz(hdr) + res) <= sender->publ.max_pkt)
1310 goto again;
1312 return link_send_sections_long(sender, msg_sect,
1313 num_sect, destaddr);
1315 tipc_node_unlock(node);
1317 read_unlock_bh(&tipc_net_lock);
1319 /* Couldn't find a link to the destination node */
1321 if (buf)
1322 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1323 if (res >= 0)
1324 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1325 TIPC_ERR_NO_NODE);
1326 return res;
1330 * link_send_sections_long(): Entry for long messages where the
1331 * destination node is known and the header is complete,
1332 * inclusive total message length.
1333 * Link and bearer congestion status have been checked to be ok,
1334 * and are ignored if they change.
1336 * Note that fragments do not use the full link MTU so that they won't have
1337 * to undergo refragmentation if link changeover causes them to be sent
1338 * over another link with an additional tunnel header added as prefix.
1339 * (Refragmentation will still occur if the other link has a smaller MTU.)
1341 * Returns user data length or errno.
1343 static int link_send_sections_long(struct port *sender,
1344 struct iovec const *msg_sect,
1345 u32 num_sect,
1346 u32 destaddr)
1348 struct link *l_ptr;
1349 struct node *node;
1350 struct tipc_msg *hdr = &sender->publ.phdr;
1351 u32 dsz = msg_data_sz(hdr);
1352 u32 max_pkt,fragm_sz,rest;
1353 struct tipc_msg fragm_hdr;
1354 struct sk_buff *buf,*buf_chain,*prev;
1355 u32 fragm_crs,fragm_rest,hsz,sect_rest;
1356 const unchar *sect_crs;
1357 int curr_sect;
1358 u32 fragm_no;
1360 again:
1361 fragm_no = 1;
1362 max_pkt = sender->publ.max_pkt - INT_H_SIZE;
1363 /* leave room for tunnel header in case of link changeover */
1364 fragm_sz = max_pkt - INT_H_SIZE;
1365 /* leave room for fragmentation header in each fragment */
1366 rest = dsz;
1367 fragm_crs = 0;
1368 fragm_rest = 0;
1369 sect_rest = 0;
1370 sect_crs = NULL;
1371 curr_sect = -1;
1373 /* Prepare reusable fragment header: */
1375 msg_dbg(hdr, ">FRAGMENTING>");
1376 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1377 TIPC_OK, INT_H_SIZE, msg_destnode(hdr));
1378 msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1379 msg_set_size(&fragm_hdr, max_pkt);
1380 msg_set_fragm_no(&fragm_hdr, 1);
1382 /* Prepare header of first fragment: */
1384 buf_chain = buf = buf_acquire(max_pkt);
1385 if (!buf)
1386 return -ENOMEM;
1387 buf->next = NULL;
1388 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1389 hsz = msg_hdr_sz(hdr);
1390 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1391 msg_dbg(buf_msg(buf), ">BUILD>");
1393 /* Chop up message: */
1395 fragm_crs = INT_H_SIZE + hsz;
1396 fragm_rest = fragm_sz - hsz;
1398 do { /* For all sections */
1399 u32 sz;
1401 if (!sect_rest) {
1402 sect_rest = msg_sect[++curr_sect].iov_len;
1403 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1406 if (sect_rest < fragm_rest)
1407 sz = sect_rest;
1408 else
1409 sz = fragm_rest;
1411 if (likely(!sender->user_port)) {
1412 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1413 error:
1414 for (; buf_chain; buf_chain = buf) {
1415 buf = buf_chain->next;
1416 buf_discard(buf_chain);
1418 return -EFAULT;
1420 } else
1421 skb_copy_to_linear_data_offset(buf, fragm_crs,
1422 sect_crs, sz);
1423 sect_crs += sz;
1424 sect_rest -= sz;
1425 fragm_crs += sz;
1426 fragm_rest -= sz;
1427 rest -= sz;
1429 if (!fragm_rest && rest) {
1431 /* Initiate new fragment: */
1432 if (rest <= fragm_sz) {
1433 fragm_sz = rest;
1434 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1435 } else {
1436 msg_set_type(&fragm_hdr, FRAGMENT);
1438 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1439 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1440 prev = buf;
1441 buf = buf_acquire(fragm_sz + INT_H_SIZE);
1442 if (!buf)
1443 goto error;
1445 buf->next = NULL;
1446 prev->next = buf;
1447 skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1448 fragm_crs = INT_H_SIZE;
1449 fragm_rest = fragm_sz;
1450 msg_dbg(buf_msg(buf)," >BUILD>");
1453 while (rest > 0);
1456 * Now we have a buffer chain. Select a link and check
1457 * that packet size is still OK
1459 node = tipc_node_select(destaddr, sender->publ.ref & 1);
1460 if (likely(node)) {
1461 tipc_node_lock(node);
1462 l_ptr = node->active_links[sender->publ.ref & 1];
1463 if (!l_ptr) {
1464 tipc_node_unlock(node);
1465 goto reject;
1467 if (link_max_pkt(l_ptr) < max_pkt) {
1468 sender->publ.max_pkt = link_max_pkt(l_ptr);
1469 tipc_node_unlock(node);
1470 for (; buf_chain; buf_chain = buf) {
1471 buf = buf_chain->next;
1472 buf_discard(buf_chain);
1474 goto again;
1476 } else {
1477 reject:
1478 for (; buf_chain; buf_chain = buf) {
1479 buf = buf_chain->next;
1480 buf_discard(buf_chain);
1482 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1483 TIPC_ERR_NO_NODE);
1486 /* Append whole chain to send queue: */
1488 buf = buf_chain;
1489 l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1490 if (!l_ptr->next_out)
1491 l_ptr->next_out = buf_chain;
1492 l_ptr->stats.sent_fragmented++;
1493 while (buf) {
1494 struct sk_buff *next = buf->next;
1495 struct tipc_msg *msg = buf_msg(buf);
1497 l_ptr->stats.sent_fragments++;
1498 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1499 link_add_to_outqueue(l_ptr, buf, msg);
1500 msg_dbg(msg, ">ADD>");
1501 buf = next;
1504 /* Send it, if possible: */
1506 tipc_link_push_queue(l_ptr);
1507 tipc_node_unlock(node);
1508 return dsz;
1512 * tipc_link_push_packet: Push one unsent packet to the media
1514 u32 tipc_link_push_packet(struct link *l_ptr)
1516 struct sk_buff *buf = l_ptr->first_out;
1517 u32 r_q_size = l_ptr->retransm_queue_size;
1518 u32 r_q_head = l_ptr->retransm_queue_head;
1520 /* Step to position where retransmission failed, if any, */
1521 /* consider that buffers may have been released in meantime */
1523 if (r_q_size && buf) {
1524 u32 last = lesser(mod(r_q_head + r_q_size),
1525 link_last_sent(l_ptr));
1526 u32 first = msg_seqno(buf_msg(buf));
1528 while (buf && less(first, r_q_head)) {
1529 first = mod(first + 1);
1530 buf = buf->next;
1532 l_ptr->retransm_queue_head = r_q_head = first;
1533 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1536 /* Continue retransmission now, if there is anything: */
1538 if (r_q_size && buf && !skb_cloned(buf)) {
1539 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1540 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1541 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1542 msg_dbg(buf_msg(buf), ">DEF-RETR>");
1543 l_ptr->retransm_queue_head = mod(++r_q_head);
1544 l_ptr->retransm_queue_size = --r_q_size;
1545 l_ptr->stats.retransmitted++;
1546 return TIPC_OK;
1547 } else {
1548 l_ptr->stats.bearer_congs++;
1549 msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1550 return PUSH_FAILED;
1554 /* Send deferred protocol message, if any: */
1556 buf = l_ptr->proto_msg_queue;
1557 if (buf) {
1558 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1559 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1560 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1561 msg_dbg(buf_msg(buf), ">DEF-PROT>");
1562 l_ptr->unacked_window = 0;
1563 buf_discard(buf);
1564 l_ptr->proto_msg_queue = NULL;
1565 return TIPC_OK;
1566 } else {
1567 msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1568 l_ptr->stats.bearer_congs++;
1569 return PUSH_FAILED;
1573 /* Send one deferred data message, if send window not full: */
1575 buf = l_ptr->next_out;
1576 if (buf) {
1577 struct tipc_msg *msg = buf_msg(buf);
1578 u32 next = msg_seqno(msg);
1579 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1581 if (mod(next - first) < l_ptr->queue_limit[0]) {
1582 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1583 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1584 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1585 if (msg_user(msg) == MSG_BUNDLER)
1586 msg_set_type(msg, CLOSED_MSG);
1587 msg_dbg(msg, ">PUSH-DATA>");
1588 l_ptr->next_out = buf->next;
1589 return TIPC_OK;
1590 } else {
1591 msg_dbg(msg, "|PUSH-DATA|");
1592 l_ptr->stats.bearer_congs++;
1593 return PUSH_FAILED;
1597 return PUSH_FINISHED;
1601 * push_queue(): push out the unsent messages of a link where
1602 * congestion has abated. Node is locked
1604 void tipc_link_push_queue(struct link *l_ptr)
1606 u32 res;
1608 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1609 return;
1611 do {
1612 res = tipc_link_push_packet(l_ptr);
1614 while (res == TIPC_OK);
1615 if (res == PUSH_FAILED)
1616 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1619 static void link_reset_all(unsigned long addr)
1621 struct node *n_ptr;
1622 char addr_string[16];
1623 u32 i;
1625 read_lock_bh(&tipc_net_lock);
1626 n_ptr = tipc_node_find((u32)addr);
1627 if (!n_ptr) {
1628 read_unlock_bh(&tipc_net_lock);
1629 return; /* node no longer exists */
1632 tipc_node_lock(n_ptr);
1634 warn("Resetting all links to %s\n",
1635 addr_string_fill(addr_string, n_ptr->addr));
1637 for (i = 0; i < MAX_BEARERS; i++) {
1638 if (n_ptr->links[i]) {
1639 link_print(n_ptr->links[i], TIPC_OUTPUT,
1640 "Resetting link\n");
1641 tipc_link_reset(n_ptr->links[i]);
1645 tipc_node_unlock(n_ptr);
1646 read_unlock_bh(&tipc_net_lock);
1649 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1651 struct tipc_msg *msg = buf_msg(buf);
1653 warn("Retransmission failure on link <%s>\n", l_ptr->name);
1654 tipc_msg_print(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1656 if (l_ptr->addr) {
1658 /* Handle failure on standard link */
1660 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1661 tipc_link_reset(l_ptr);
1663 } else {
1665 /* Handle failure on broadcast link */
1667 struct node *n_ptr;
1668 char addr_string[16];
1670 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u, ", msg_seqno(msg));
1671 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %lu\n",
1672 (unsigned long) TIPC_SKB_CB(buf)->handle);
1674 n_ptr = l_ptr->owner->next;
1675 tipc_node_lock(n_ptr);
1677 addr_string_fill(addr_string, n_ptr->addr);
1678 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1679 tipc_printf(TIPC_OUTPUT, "Supported: %d, ", n_ptr->bclink.supported);
1680 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1681 tipc_printf(TIPC_OUTPUT, "Last in: %u, ", n_ptr->bclink.last_in);
1682 tipc_printf(TIPC_OUTPUT, "Gap after: %u, ", n_ptr->bclink.gap_after);
1683 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1684 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1686 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1688 tipc_node_unlock(n_ptr);
1690 l_ptr->stale_count = 0;
1694 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1695 u32 retransmits)
1697 struct tipc_msg *msg;
1699 if (!buf)
1700 return;
1702 msg = buf_msg(buf);
1704 dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1706 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1707 if (!skb_cloned(buf)) {
1708 msg_dbg(msg, ">NO_RETR->BCONG>");
1709 dbg_print_link(l_ptr, " ");
1710 l_ptr->retransm_queue_head = msg_seqno(msg);
1711 l_ptr->retransm_queue_size = retransmits;
1712 return;
1713 } else {
1714 /* Don't retransmit if driver already has the buffer */
1716 } else {
1717 /* Detect repeated retransmit failures on uncongested bearer */
1719 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1720 if (++l_ptr->stale_count > 100) {
1721 link_retransmit_failure(l_ptr, buf);
1722 return;
1724 } else {
1725 l_ptr->last_retransmitted = msg_seqno(msg);
1726 l_ptr->stale_count = 1;
1730 while (retransmits && (buf != l_ptr->next_out) && buf && !skb_cloned(buf)) {
1731 msg = buf_msg(buf);
1732 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1733 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1734 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1735 msg_dbg(buf_msg(buf), ">RETR>");
1736 buf = buf->next;
1737 retransmits--;
1738 l_ptr->stats.retransmitted++;
1739 } else {
1740 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1741 l_ptr->stats.bearer_congs++;
1742 l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1743 l_ptr->retransm_queue_size = retransmits;
1744 return;
1748 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1752 * link_recv_non_seq: Receive packets which are outside
1753 * the link sequence flow
1756 static void link_recv_non_seq(struct sk_buff *buf)
1758 struct tipc_msg *msg = buf_msg(buf);
1760 if (msg_user(msg) == LINK_CONFIG)
1761 tipc_disc_recv_msg(buf);
1762 else
1763 tipc_bclink_recv_pkt(buf);
1767 * link_insert_deferred_queue - insert deferred messages back into receive chain
1770 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1771 struct sk_buff *buf)
1773 u32 seq_no;
1775 if (l_ptr->oldest_deferred_in == NULL)
1776 return buf;
1778 seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1779 if (seq_no == mod(l_ptr->next_in_no)) {
1780 l_ptr->newest_deferred_in->next = buf;
1781 buf = l_ptr->oldest_deferred_in;
1782 l_ptr->oldest_deferred_in = NULL;
1783 l_ptr->deferred_inqueue_sz = 0;
1785 return buf;
1788 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1790 read_lock_bh(&tipc_net_lock);
1791 while (head) {
1792 struct bearer *b_ptr;
1793 struct node *n_ptr;
1794 struct link *l_ptr;
1795 struct sk_buff *crs;
1796 struct sk_buff *buf = head;
1797 struct tipc_msg *msg = buf_msg(buf);
1798 u32 seq_no = msg_seqno(msg);
1799 u32 ackd = msg_ack(msg);
1800 u32 released = 0;
1801 int type;
1803 b_ptr = (struct bearer *)tb_ptr;
1804 TIPC_SKB_CB(buf)->handle = b_ptr;
1806 head = head->next;
1807 if (unlikely(msg_version(msg) != TIPC_VERSION))
1808 goto cont;
1809 #if 0
1810 if (msg_user(msg) != LINK_PROTOCOL)
1811 #endif
1812 msg_dbg(msg,"<REC<");
1814 if (unlikely(msg_non_seq(msg))) {
1815 link_recv_non_seq(buf);
1816 continue;
1819 if (unlikely(!msg_short(msg) &&
1820 (msg_destnode(msg) != tipc_own_addr)))
1821 goto cont;
1823 n_ptr = tipc_node_find(msg_prevnode(msg));
1824 if (unlikely(!n_ptr))
1825 goto cont;
1827 tipc_node_lock(n_ptr);
1828 l_ptr = n_ptr->links[b_ptr->identity];
1829 if (unlikely(!l_ptr)) {
1830 tipc_node_unlock(n_ptr);
1831 goto cont;
1834 * Release acked messages
1836 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1837 if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1838 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1841 crs = l_ptr->first_out;
1842 while ((crs != l_ptr->next_out) &&
1843 less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1844 struct sk_buff *next = crs->next;
1846 buf_discard(crs);
1847 crs = next;
1848 released++;
1850 if (released) {
1851 l_ptr->first_out = crs;
1852 l_ptr->out_queue_size -= released;
1854 if (unlikely(l_ptr->next_out))
1855 tipc_link_push_queue(l_ptr);
1856 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1857 tipc_link_wakeup_ports(l_ptr, 0);
1858 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1859 l_ptr->stats.sent_acks++;
1860 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1863 protocol_check:
1864 if (likely(link_working_working(l_ptr))) {
1865 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1866 l_ptr->next_in_no++;
1867 if (unlikely(l_ptr->oldest_deferred_in))
1868 head = link_insert_deferred_queue(l_ptr,
1869 head);
1870 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1871 deliver:
1872 if (likely(msg_isdata(msg))) {
1873 tipc_node_unlock(n_ptr);
1874 tipc_port_recv_msg(buf);
1875 continue;
1877 switch (msg_user(msg)) {
1878 case MSG_BUNDLER:
1879 l_ptr->stats.recv_bundles++;
1880 l_ptr->stats.recv_bundled +=
1881 msg_msgcnt(msg);
1882 tipc_node_unlock(n_ptr);
1883 tipc_link_recv_bundle(buf);
1884 continue;
1885 case ROUTE_DISTRIBUTOR:
1886 tipc_node_unlock(n_ptr);
1887 tipc_cltr_recv_routing_table(buf);
1888 continue;
1889 case NAME_DISTRIBUTOR:
1890 tipc_node_unlock(n_ptr);
1891 tipc_named_recv(buf);
1892 continue;
1893 case CONN_MANAGER:
1894 tipc_node_unlock(n_ptr);
1895 tipc_port_recv_proto_msg(buf);
1896 continue;
1897 case MSG_FRAGMENTER:
1898 l_ptr->stats.recv_fragments++;
1899 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1900 &buf, &msg)) {
1901 l_ptr->stats.recv_fragmented++;
1902 goto deliver;
1904 break;
1905 case CHANGEOVER_PROTOCOL:
1906 type = msg_type(msg);
1907 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1908 msg = buf_msg(buf);
1909 seq_no = msg_seqno(msg);
1910 TIPC_SKB_CB(buf)->handle
1911 = b_ptr;
1912 if (type == ORIGINAL_MSG)
1913 goto deliver;
1914 goto protocol_check;
1916 break;
1919 tipc_node_unlock(n_ptr);
1920 tipc_net_route_msg(buf);
1921 continue;
1923 link_handle_out_of_seq_msg(l_ptr, buf);
1924 head = link_insert_deferred_queue(l_ptr, head);
1925 tipc_node_unlock(n_ptr);
1926 continue;
1929 if (msg_user(msg) == LINK_PROTOCOL) {
1930 link_recv_proto_msg(l_ptr, buf);
1931 head = link_insert_deferred_queue(l_ptr, head);
1932 tipc_node_unlock(n_ptr);
1933 continue;
1935 msg_dbg(msg,"NSEQ<REC<");
1936 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1938 if (link_working_working(l_ptr)) {
1939 /* Re-insert in front of queue */
1940 msg_dbg(msg,"RECV-REINS:");
1941 buf->next = head;
1942 head = buf;
1943 tipc_node_unlock(n_ptr);
1944 continue;
1946 tipc_node_unlock(n_ptr);
1947 cont:
1948 buf_discard(buf);
1950 read_unlock_bh(&tipc_net_lock);
1954 * link_defer_buf(): Sort a received out-of-sequence packet
1955 * into the deferred reception queue.
1956 * Returns the increase of the queue length,i.e. 0 or 1
1959 u32 tipc_link_defer_pkt(struct sk_buff **head,
1960 struct sk_buff **tail,
1961 struct sk_buff *buf)
1963 struct sk_buff *prev = NULL;
1964 struct sk_buff *crs = *head;
1965 u32 seq_no = msg_seqno(buf_msg(buf));
1967 buf->next = NULL;
1969 /* Empty queue ? */
1970 if (*head == NULL) {
1971 *head = *tail = buf;
1972 return 1;
1975 /* Last ? */
1976 if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1977 (*tail)->next = buf;
1978 *tail = buf;
1979 return 1;
1982 /* Scan through queue and sort it in */
1983 do {
1984 struct tipc_msg *msg = buf_msg(crs);
1986 if (less(seq_no, msg_seqno(msg))) {
1987 buf->next = crs;
1988 if (prev)
1989 prev->next = buf;
1990 else
1991 *head = buf;
1992 return 1;
1994 if (seq_no == msg_seqno(msg)) {
1995 break;
1997 prev = crs;
1998 crs = crs->next;
2000 while (crs);
2002 /* Message is a duplicate of an existing message */
2004 buf_discard(buf);
2005 return 0;
2009 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2012 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2013 struct sk_buff *buf)
2015 u32 seq_no = msg_seqno(buf_msg(buf));
2017 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2018 link_recv_proto_msg(l_ptr, buf);
2019 return;
2022 dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2023 seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2025 /* Record OOS packet arrival (force mismatch on next timeout) */
2027 l_ptr->checkpoint--;
2030 * Discard packet if a duplicate; otherwise add it to deferred queue
2031 * and notify peer of gap as per protocol specification
2034 if (less(seq_no, mod(l_ptr->next_in_no))) {
2035 l_ptr->stats.duplicates++;
2036 buf_discard(buf);
2037 return;
2040 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2041 &l_ptr->newest_deferred_in, buf)) {
2042 l_ptr->deferred_inqueue_sz++;
2043 l_ptr->stats.deferred_recv++;
2044 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2045 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2046 } else
2047 l_ptr->stats.duplicates++;
2051 * Send protocol message to the other endpoint.
2053 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2054 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2056 struct sk_buff *buf = NULL;
2057 struct tipc_msg *msg = l_ptr->pmsg;
2058 u32 msg_size = sizeof(l_ptr->proto_msg);
2060 if (link_blocked(l_ptr))
2061 return;
2062 msg_set_type(msg, msg_typ);
2063 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2064 msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2065 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2067 if (msg_typ == STATE_MSG) {
2068 u32 next_sent = mod(l_ptr->next_out_no);
2070 if (!tipc_link_is_up(l_ptr))
2071 return;
2072 if (l_ptr->next_out)
2073 next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2074 msg_set_next_sent(msg, next_sent);
2075 if (l_ptr->oldest_deferred_in) {
2076 u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2077 gap = mod(rec - mod(l_ptr->next_in_no));
2079 msg_set_seq_gap(msg, gap);
2080 if (gap)
2081 l_ptr->stats.sent_nacks++;
2082 msg_set_link_tolerance(msg, tolerance);
2083 msg_set_linkprio(msg, priority);
2084 msg_set_max_pkt(msg, ack_mtu);
2085 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2086 msg_set_probe(msg, probe_msg != 0);
2087 if (probe_msg) {
2088 u32 mtu = l_ptr->max_pkt;
2090 if ((mtu < l_ptr->max_pkt_target) &&
2091 link_working_working(l_ptr) &&
2092 l_ptr->fsm_msg_cnt) {
2093 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2094 if (l_ptr->max_pkt_probes == 10) {
2095 l_ptr->max_pkt_target = (msg_size - 4);
2096 l_ptr->max_pkt_probes = 0;
2097 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2099 l_ptr->max_pkt_probes++;
2102 l_ptr->stats.sent_probes++;
2104 l_ptr->stats.sent_states++;
2105 } else { /* RESET_MSG or ACTIVATE_MSG */
2106 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2107 msg_set_seq_gap(msg, 0);
2108 msg_set_next_sent(msg, 1);
2109 msg_set_link_tolerance(msg, l_ptr->tolerance);
2110 msg_set_linkprio(msg, l_ptr->priority);
2111 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2114 if (tipc_node_has_redundant_links(l_ptr->owner)) {
2115 msg_set_redundant_link(msg);
2116 } else {
2117 msg_clear_redundant_link(msg);
2119 msg_set_linkprio(msg, l_ptr->priority);
2121 /* Ensure sequence number will not fit : */
2123 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2125 /* Congestion? */
2127 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2128 if (!l_ptr->proto_msg_queue) {
2129 l_ptr->proto_msg_queue =
2130 buf_acquire(sizeof(l_ptr->proto_msg));
2132 buf = l_ptr->proto_msg_queue;
2133 if (!buf)
2134 return;
2135 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2136 return;
2138 msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2140 /* Message can be sent */
2142 msg_dbg(msg, ">>");
2144 buf = buf_acquire(msg_size);
2145 if (!buf)
2146 return;
2148 skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
2149 msg_set_size(buf_msg(buf), msg_size);
2151 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2152 l_ptr->unacked_window = 0;
2153 buf_discard(buf);
2154 return;
2157 /* New congestion */
2158 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2159 l_ptr->proto_msg_queue = buf;
2160 l_ptr->stats.bearer_congs++;
2164 * Receive protocol message :
2165 * Note that network plane id propagates through the network, and may
2166 * change at any time. The node with lowest address rules
2169 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2171 u32 rec_gap = 0;
2172 u32 max_pkt_info;
2173 u32 max_pkt_ack;
2174 u32 msg_tol;
2175 struct tipc_msg *msg = buf_msg(buf);
2177 dbg("AT(%u):", jiffies_to_msecs(jiffies));
2178 msg_dbg(msg, "<<");
2179 if (link_blocked(l_ptr))
2180 goto exit;
2182 /* record unnumbered packet arrival (force mismatch on next timeout) */
2184 l_ptr->checkpoint--;
2186 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2187 if (tipc_own_addr > msg_prevnode(msg))
2188 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2190 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2192 switch (msg_type(msg)) {
2194 case RESET_MSG:
2195 if (!link_working_unknown(l_ptr) && l_ptr->peer_session) {
2196 if (msg_session(msg) == l_ptr->peer_session) {
2197 dbg("Duplicate RESET: %u<->%u\n",
2198 msg_session(msg), l_ptr->peer_session);
2199 break; /* duplicate: ignore */
2202 /* fall thru' */
2203 case ACTIVATE_MSG:
2204 /* Update link settings according other endpoint's values */
2206 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2208 if ((msg_tol = msg_link_tolerance(msg)) &&
2209 (msg_tol > l_ptr->tolerance))
2210 link_set_supervision_props(l_ptr, msg_tol);
2212 if (msg_linkprio(msg) > l_ptr->priority)
2213 l_ptr->priority = msg_linkprio(msg);
2215 max_pkt_info = msg_max_pkt(msg);
2216 if (max_pkt_info) {
2217 if (max_pkt_info < l_ptr->max_pkt_target)
2218 l_ptr->max_pkt_target = max_pkt_info;
2219 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2220 l_ptr->max_pkt = l_ptr->max_pkt_target;
2221 } else {
2222 l_ptr->max_pkt = l_ptr->max_pkt_target;
2224 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2226 link_state_event(l_ptr, msg_type(msg));
2228 l_ptr->peer_session = msg_session(msg);
2229 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2231 /* Synchronize broadcast sequence numbers */
2232 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2233 l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2235 break;
2236 case STATE_MSG:
2238 if ((msg_tol = msg_link_tolerance(msg)))
2239 link_set_supervision_props(l_ptr, msg_tol);
2241 if (msg_linkprio(msg) &&
2242 (msg_linkprio(msg) != l_ptr->priority)) {
2243 warn("Resetting link <%s>, priority change %u->%u\n",
2244 l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2245 l_ptr->priority = msg_linkprio(msg);
2246 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2247 break;
2249 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2250 l_ptr->stats.recv_states++;
2251 if (link_reset_unknown(l_ptr))
2252 break;
2254 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2255 rec_gap = mod(msg_next_sent(msg) -
2256 mod(l_ptr->next_in_no));
2259 max_pkt_ack = msg_max_pkt(msg);
2260 if (max_pkt_ack > l_ptr->max_pkt) {
2261 dbg("Link <%s> updated MTU %u -> %u\n",
2262 l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2263 l_ptr->max_pkt = max_pkt_ack;
2264 l_ptr->max_pkt_probes = 0;
2267 max_pkt_ack = 0;
2268 if (msg_probe(msg)) {
2269 l_ptr->stats.recv_probes++;
2270 if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2271 max_pkt_ack = msg_size(msg);
2275 /* Protocol message before retransmits, reduce loss risk */
2277 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2279 if (rec_gap || (msg_probe(msg))) {
2280 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2281 0, rec_gap, 0, 0, max_pkt_ack);
2283 if (msg_seq_gap(msg)) {
2284 msg_dbg(msg, "With Gap:");
2285 l_ptr->stats.recv_nacks++;
2286 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2287 msg_seq_gap(msg));
2289 break;
2290 default:
2291 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2293 exit:
2294 buf_discard(buf);
2299 * tipc_link_tunnel(): Send one message via a link belonging to
2300 * another bearer. Owner node is locked.
2302 void tipc_link_tunnel(struct link *l_ptr,
2303 struct tipc_msg *tunnel_hdr,
2304 struct tipc_msg *msg,
2305 u32 selector)
2307 struct link *tunnel;
2308 struct sk_buff *buf;
2309 u32 length = msg_size(msg);
2311 tunnel = l_ptr->owner->active_links[selector & 1];
2312 if (!tipc_link_is_up(tunnel)) {
2313 warn("Link changeover error, "
2314 "tunnel link no longer available\n");
2315 return;
2317 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2318 buf = buf_acquire(length + INT_H_SIZE);
2319 if (!buf) {
2320 warn("Link changeover error, "
2321 "unable to send tunnel msg\n");
2322 return;
2324 skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
2325 skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
2326 dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2327 msg_dbg(buf_msg(buf), ">SEND>");
2328 tipc_link_send_buf(tunnel, buf);
2334 * changeover(): Send whole message queue via the remaining link
2335 * Owner node is locked.
2338 void tipc_link_changeover(struct link *l_ptr)
2340 u32 msgcount = l_ptr->out_queue_size;
2341 struct sk_buff *crs = l_ptr->first_out;
2342 struct link *tunnel = l_ptr->owner->active_links[0];
2343 struct tipc_msg tunnel_hdr;
2344 int split_bundles;
2346 if (!tunnel)
2347 return;
2349 if (!l_ptr->owner->permit_changeover) {
2350 warn("Link changeover error, "
2351 "peer did not permit changeover\n");
2352 return;
2355 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2356 ORIGINAL_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2357 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2358 msg_set_msgcnt(&tunnel_hdr, msgcount);
2359 dbg("Link changeover requires %u tunnel messages\n", msgcount);
2361 if (!l_ptr->first_out) {
2362 struct sk_buff *buf;
2364 buf = buf_acquire(INT_H_SIZE);
2365 if (buf) {
2366 skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2367 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2368 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2369 tunnel->b_ptr->net_plane);
2370 msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2371 tipc_link_send_buf(tunnel, buf);
2372 } else {
2373 warn("Link changeover error, "
2374 "unable to send changeover msg\n");
2376 return;
2379 split_bundles = (l_ptr->owner->active_links[0] !=
2380 l_ptr->owner->active_links[1]);
2382 while (crs) {
2383 struct tipc_msg *msg = buf_msg(crs);
2385 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2386 struct tipc_msg *m = msg_get_wrapped(msg);
2387 unchar* pos = (unchar*)m;
2389 msgcount = msg_msgcnt(msg);
2390 while (msgcount--) {
2391 msg_set_seqno(m,msg_seqno(msg));
2392 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2393 msg_link_selector(m));
2394 pos += align(msg_size(m));
2395 m = (struct tipc_msg *)pos;
2397 } else {
2398 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2399 msg_link_selector(msg));
2401 crs = crs->next;
2405 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2407 struct sk_buff *iter;
2408 struct tipc_msg tunnel_hdr;
2410 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2411 DUPLICATE_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2412 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2413 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2414 iter = l_ptr->first_out;
2415 while (iter) {
2416 struct sk_buff *outbuf;
2417 struct tipc_msg *msg = buf_msg(iter);
2418 u32 length = msg_size(msg);
2420 if (msg_user(msg) == MSG_BUNDLER)
2421 msg_set_type(msg, CLOSED_MSG);
2422 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2423 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2424 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2425 outbuf = buf_acquire(length + INT_H_SIZE);
2426 if (outbuf == NULL) {
2427 warn("Link changeover error, "
2428 "unable to send duplicate msg\n");
2429 return;
2431 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2432 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2433 length);
2434 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2435 tunnel->b_ptr->net_plane);
2436 msg_dbg(buf_msg(outbuf), ">SEND>");
2437 tipc_link_send_buf(tunnel, outbuf);
2438 if (!tipc_link_is_up(l_ptr))
2439 return;
2440 iter = iter->next;
2447 * buf_extract - extracts embedded TIPC message from another message
2448 * @skb: encapsulating message buffer
2449 * @from_pos: offset to extract from
2451 * Returns a new message buffer containing an embedded message. The
2452 * encapsulating message itself is left unchanged.
2455 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2457 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2458 u32 size = msg_size(msg);
2459 struct sk_buff *eb;
2461 eb = buf_acquire(size);
2462 if (eb)
2463 skb_copy_to_linear_data(eb, msg, size);
2464 return eb;
2468 * link_recv_changeover_msg(): Receive tunneled packet sent
2469 * via other link. Node is locked. Return extracted buffer.
2472 static int link_recv_changeover_msg(struct link **l_ptr,
2473 struct sk_buff **buf)
2475 struct sk_buff *tunnel_buf = *buf;
2476 struct link *dest_link;
2477 struct tipc_msg *msg;
2478 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2479 u32 msg_typ = msg_type(tunnel_msg);
2480 u32 msg_count = msg_msgcnt(tunnel_msg);
2482 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2483 if (!dest_link) {
2484 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2485 goto exit;
2487 if (dest_link == *l_ptr) {
2488 err("Unexpected changeover message on link <%s>\n",
2489 (*l_ptr)->name);
2490 goto exit;
2492 dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2493 (*l_ptr)->b_ptr->net_plane);
2494 *l_ptr = dest_link;
2495 msg = msg_get_wrapped(tunnel_msg);
2497 if (msg_typ == DUPLICATE_MSG) {
2498 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2499 msg_dbg(tunnel_msg, "DROP/<REC<");
2500 goto exit;
2502 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2503 if (*buf == NULL) {
2504 warn("Link changeover error, duplicate msg dropped\n");
2505 goto exit;
2507 msg_dbg(tunnel_msg, "TNL<REC<");
2508 buf_discard(tunnel_buf);
2509 return 1;
2512 /* First original message ?: */
2514 if (tipc_link_is_up(dest_link)) {
2515 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2516 info("Resetting link <%s>, changeover initiated by peer\n",
2517 dest_link->name);
2518 tipc_link_reset(dest_link);
2519 dest_link->exp_msg_count = msg_count;
2520 dbg("Expecting %u tunnelled messages\n", msg_count);
2521 if (!msg_count)
2522 goto exit;
2523 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2524 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2525 dest_link->exp_msg_count = msg_count;
2526 dbg("Expecting %u tunnelled messages\n", msg_count);
2527 if (!msg_count)
2528 goto exit;
2531 /* Receive original message */
2533 if (dest_link->exp_msg_count == 0) {
2534 warn("Link switchover error, "
2535 "got too many tunnelled messages\n");
2536 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2537 dbg_print_link(dest_link, "LINK:");
2538 goto exit;
2540 dest_link->exp_msg_count--;
2541 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2542 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2543 goto exit;
2544 } else {
2545 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2546 if (*buf != NULL) {
2547 msg_dbg(tunnel_msg, "TNL<REC<");
2548 buf_discard(tunnel_buf);
2549 return 1;
2550 } else {
2551 warn("Link changeover error, original msg dropped\n");
2554 exit:
2555 *buf = NULL;
2556 buf_discard(tunnel_buf);
2557 return 0;
2561 * Bundler functionality:
2563 void tipc_link_recv_bundle(struct sk_buff *buf)
2565 u32 msgcount = msg_msgcnt(buf_msg(buf));
2566 u32 pos = INT_H_SIZE;
2567 struct sk_buff *obuf;
2569 msg_dbg(buf_msg(buf), "<BNDL<: ");
2570 while (msgcount--) {
2571 obuf = buf_extract(buf, pos);
2572 if (obuf == NULL) {
2573 warn("Link unable to unbundle message(s)\n");
2574 break;
2576 pos += align(msg_size(buf_msg(obuf)));
2577 msg_dbg(buf_msg(obuf), " /");
2578 tipc_net_route_msg(obuf);
2580 buf_discard(buf);
2584 * Fragmentation/defragmentation:
2589 * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2590 * The buffer is complete, inclusive total message length.
2591 * Returns user data length.
2593 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2595 struct tipc_msg *inmsg = buf_msg(buf);
2596 struct tipc_msg fragm_hdr;
2597 u32 insize = msg_size(inmsg);
2598 u32 dsz = msg_data_sz(inmsg);
2599 unchar *crs = buf->data;
2600 u32 rest = insize;
2601 u32 pack_sz = link_max_pkt(l_ptr);
2602 u32 fragm_sz = pack_sz - INT_H_SIZE;
2603 u32 fragm_no = 1;
2604 u32 destaddr = msg_destnode(inmsg);
2606 if (msg_short(inmsg))
2607 destaddr = l_ptr->addr;
2609 if (msg_routed(inmsg))
2610 msg_set_prevnode(inmsg, tipc_own_addr);
2612 /* Prepare reusable fragment header: */
2614 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2615 TIPC_OK, INT_H_SIZE, destaddr);
2616 msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2617 msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2618 msg_set_fragm_no(&fragm_hdr, fragm_no);
2619 l_ptr->stats.sent_fragmented++;
2621 /* Chop up message: */
2623 while (rest > 0) {
2624 struct sk_buff *fragm;
2626 if (rest <= fragm_sz) {
2627 fragm_sz = rest;
2628 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2630 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2631 if (fragm == NULL) {
2632 warn("Link unable to fragment message\n");
2633 dsz = -ENOMEM;
2634 goto exit;
2636 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2637 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2638 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2639 fragm_sz);
2640 /* Send queued messages first, if any: */
2642 l_ptr->stats.sent_fragments++;
2643 tipc_link_send_buf(l_ptr, fragm);
2644 if (!tipc_link_is_up(l_ptr))
2645 return dsz;
2646 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2647 rest -= fragm_sz;
2648 crs += fragm_sz;
2649 msg_set_type(&fragm_hdr, FRAGMENT);
2651 exit:
2652 buf_discard(buf);
2653 return dsz;
2657 * A pending message being re-assembled must store certain values
2658 * to handle subsequent fragments correctly. The following functions
2659 * help storing these values in unused, available fields in the
2660 * pending message. This makes dynamic memory allocation unecessary.
2663 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2665 msg_set_seqno(buf_msg(buf), seqno);
2668 static u32 get_fragm_size(struct sk_buff *buf)
2670 return msg_ack(buf_msg(buf));
2673 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2675 msg_set_ack(buf_msg(buf), sz);
2678 static u32 get_expected_frags(struct sk_buff *buf)
2680 return msg_bcast_ack(buf_msg(buf));
2683 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2685 msg_set_bcast_ack(buf_msg(buf), exp);
2688 static u32 get_timer_cnt(struct sk_buff *buf)
2690 return msg_reroute_cnt(buf_msg(buf));
2693 static void incr_timer_cnt(struct sk_buff *buf)
2695 msg_incr_reroute_cnt(buf_msg(buf));
2699 * tipc_link_recv_fragment(): Called with node lock on. Returns
2700 * the reassembled buffer if message is complete.
2702 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2703 struct tipc_msg **m)
2705 struct sk_buff *prev = NULL;
2706 struct sk_buff *fbuf = *fb;
2707 struct tipc_msg *fragm = buf_msg(fbuf);
2708 struct sk_buff *pbuf = *pending;
2709 u32 long_msg_seq_no = msg_long_msgno(fragm);
2711 *fb = NULL;
2712 msg_dbg(fragm,"FRG<REC<");
2714 /* Is there an incomplete message waiting for this fragment? */
2716 while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no)
2717 || (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2718 prev = pbuf;
2719 pbuf = pbuf->next;
2722 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2723 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2724 u32 msg_sz = msg_size(imsg);
2725 u32 fragm_sz = msg_data_sz(fragm);
2726 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2727 u32 max = TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2728 if (msg_type(imsg) == TIPC_MCAST_MSG)
2729 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2730 if (msg_size(imsg) > max) {
2731 msg_dbg(fragm,"<REC<Oversized: ");
2732 buf_discard(fbuf);
2733 return 0;
2735 pbuf = buf_acquire(msg_size(imsg));
2736 if (pbuf != NULL) {
2737 pbuf->next = *pending;
2738 *pending = pbuf;
2739 skb_copy_to_linear_data(pbuf, imsg,
2740 msg_data_sz(fragm));
2741 /* Prepare buffer for subsequent fragments. */
2743 set_long_msg_seqno(pbuf, long_msg_seq_no);
2744 set_fragm_size(pbuf,fragm_sz);
2745 set_expected_frags(pbuf,exp_fragm_cnt - 1);
2746 } else {
2747 warn("Link unable to reassemble fragmented message\n");
2749 buf_discard(fbuf);
2750 return 0;
2751 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2752 u32 dsz = msg_data_sz(fragm);
2753 u32 fsz = get_fragm_size(pbuf);
2754 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2755 u32 exp_frags = get_expected_frags(pbuf) - 1;
2756 skb_copy_to_linear_data_offset(pbuf, crs,
2757 msg_data(fragm), dsz);
2758 buf_discard(fbuf);
2760 /* Is message complete? */
2762 if (exp_frags == 0) {
2763 if (prev)
2764 prev->next = pbuf->next;
2765 else
2766 *pending = pbuf->next;
2767 msg_reset_reroute_cnt(buf_msg(pbuf));
2768 *fb = pbuf;
2769 *m = buf_msg(pbuf);
2770 return 1;
2772 set_expected_frags(pbuf,exp_frags);
2773 return 0;
2775 dbg(" Discarding orphan fragment %x\n",fbuf);
2776 msg_dbg(fragm,"ORPHAN:");
2777 dbg("Pending long buffers:\n");
2778 dbg_print_buf_chain(*pending);
2779 buf_discard(fbuf);
2780 return 0;
2784 * link_check_defragm_bufs - flush stale incoming message fragments
2785 * @l_ptr: pointer to link
2788 static void link_check_defragm_bufs(struct link *l_ptr)
2790 struct sk_buff *prev = NULL;
2791 struct sk_buff *next = NULL;
2792 struct sk_buff *buf = l_ptr->defragm_buf;
2794 if (!buf)
2795 return;
2796 if (!link_working_working(l_ptr))
2797 return;
2798 while (buf) {
2799 u32 cnt = get_timer_cnt(buf);
2801 next = buf->next;
2802 if (cnt < 4) {
2803 incr_timer_cnt(buf);
2804 prev = buf;
2805 } else {
2806 dbg(" Discarding incomplete long buffer\n");
2807 msg_dbg(buf_msg(buf), "LONG:");
2808 dbg_print_link(l_ptr, "curr:");
2809 dbg("Pending long buffers:\n");
2810 dbg_print_buf_chain(l_ptr->defragm_buf);
2811 if (prev)
2812 prev->next = buf->next;
2813 else
2814 l_ptr->defragm_buf = buf->next;
2815 buf_discard(buf);
2817 buf = next;
2823 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2825 l_ptr->tolerance = tolerance;
2826 l_ptr->continuity_interval =
2827 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2828 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2832 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2834 /* Data messages from this node, inclusive FIRST_FRAGM */
2835 l_ptr->queue_limit[DATA_LOW] = window;
2836 l_ptr->queue_limit[DATA_MEDIUM] = (window / 3) * 4;
2837 l_ptr->queue_limit[DATA_HIGH] = (window / 3) * 5;
2838 l_ptr->queue_limit[DATA_CRITICAL] = (window / 3) * 6;
2839 /* Transiting data messages,inclusive FIRST_FRAGM */
2840 l_ptr->queue_limit[DATA_LOW + 4] = 300;
2841 l_ptr->queue_limit[DATA_MEDIUM + 4] = 600;
2842 l_ptr->queue_limit[DATA_HIGH + 4] = 900;
2843 l_ptr->queue_limit[DATA_CRITICAL + 4] = 1200;
2844 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2845 l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2846 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2847 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2848 /* FRAGMENT and LAST_FRAGMENT packets */
2849 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2853 * link_find_link - locate link by name
2854 * @name - ptr to link name string
2855 * @node - ptr to area to be filled with ptr to associated node
2857 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2858 * this also prevents link deletion.
2860 * Returns pointer to link (or 0 if invalid link name).
2863 static struct link *link_find_link(const char *name, struct node **node)
2865 struct link_name link_name_parts;
2866 struct bearer *b_ptr;
2867 struct link *l_ptr;
2869 if (!link_name_validate(name, &link_name_parts))
2870 return NULL;
2872 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2873 if (!b_ptr)
2874 return NULL;
2876 *node = tipc_node_find(link_name_parts.addr_peer);
2877 if (!*node)
2878 return NULL;
2880 l_ptr = (*node)->links[b_ptr->identity];
2881 if (!l_ptr || strcmp(l_ptr->name, name))
2882 return NULL;
2884 return l_ptr;
2887 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2888 u16 cmd)
2890 struct tipc_link_config *args;
2891 u32 new_value;
2892 struct link *l_ptr;
2893 struct node *node;
2894 int res;
2896 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2897 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2899 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2900 new_value = ntohl(args->value);
2902 if (!strcmp(args->name, tipc_bclink_name)) {
2903 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2904 (tipc_bclink_set_queue_limits(new_value) == 0))
2905 return tipc_cfg_reply_none();
2906 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2907 " (cannot change setting on broadcast link)");
2910 read_lock_bh(&tipc_net_lock);
2911 l_ptr = link_find_link(args->name, &node);
2912 if (!l_ptr) {
2913 read_unlock_bh(&tipc_net_lock);
2914 return tipc_cfg_reply_error_string("link not found");
2917 tipc_node_lock(node);
2918 res = -EINVAL;
2919 switch (cmd) {
2920 case TIPC_CMD_SET_LINK_TOL:
2921 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2922 (new_value <= TIPC_MAX_LINK_TOL)) {
2923 link_set_supervision_props(l_ptr, new_value);
2924 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2925 0, 0, new_value, 0, 0);
2926 res = TIPC_OK;
2928 break;
2929 case TIPC_CMD_SET_LINK_PRI:
2930 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2931 (new_value <= TIPC_MAX_LINK_PRI)) {
2932 l_ptr->priority = new_value;
2933 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2934 0, 0, 0, new_value, 0);
2935 res = TIPC_OK;
2937 break;
2938 case TIPC_CMD_SET_LINK_WINDOW:
2939 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2940 (new_value <= TIPC_MAX_LINK_WIN)) {
2941 tipc_link_set_queue_limits(l_ptr, new_value);
2942 res = TIPC_OK;
2944 break;
2946 tipc_node_unlock(node);
2948 read_unlock_bh(&tipc_net_lock);
2949 if (res)
2950 return tipc_cfg_reply_error_string("cannot change link setting");
2952 return tipc_cfg_reply_none();
2956 * link_reset_statistics - reset link statistics
2957 * @l_ptr: pointer to link
2960 static void link_reset_statistics(struct link *l_ptr)
2962 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2963 l_ptr->stats.sent_info = l_ptr->next_out_no;
2964 l_ptr->stats.recv_info = l_ptr->next_in_no;
2967 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2969 char *link_name;
2970 struct link *l_ptr;
2971 struct node *node;
2973 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2974 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2976 link_name = (char *)TLV_DATA(req_tlv_area);
2977 if (!strcmp(link_name, tipc_bclink_name)) {
2978 if (tipc_bclink_reset_stats())
2979 return tipc_cfg_reply_error_string("link not found");
2980 return tipc_cfg_reply_none();
2983 read_lock_bh(&tipc_net_lock);
2984 l_ptr = link_find_link(link_name, &node);
2985 if (!l_ptr) {
2986 read_unlock_bh(&tipc_net_lock);
2987 return tipc_cfg_reply_error_string("link not found");
2990 tipc_node_lock(node);
2991 link_reset_statistics(l_ptr);
2992 tipc_node_unlock(node);
2993 read_unlock_bh(&tipc_net_lock);
2994 return tipc_cfg_reply_none();
2998 * percent - convert count to a percentage of total (rounding up or down)
3001 static u32 percent(u32 count, u32 total)
3003 return (count * 100 + (total / 2)) / total;
3007 * tipc_link_stats - print link statistics
3008 * @name: link name
3009 * @buf: print buffer area
3010 * @buf_size: size of print buffer area
3012 * Returns length of print buffer data string (or 0 if error)
3015 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3017 struct print_buf pb;
3018 struct link *l_ptr;
3019 struct node *node;
3020 char *status;
3021 u32 profile_total = 0;
3023 if (!strcmp(name, tipc_bclink_name))
3024 return tipc_bclink_stats(buf, buf_size);
3026 tipc_printbuf_init(&pb, buf, buf_size);
3028 read_lock_bh(&tipc_net_lock);
3029 l_ptr = link_find_link(name, &node);
3030 if (!l_ptr) {
3031 read_unlock_bh(&tipc_net_lock);
3032 return 0;
3034 tipc_node_lock(node);
3036 if (tipc_link_is_active(l_ptr))
3037 status = "ACTIVE";
3038 else if (tipc_link_is_up(l_ptr))
3039 status = "STANDBY";
3040 else
3041 status = "DEFUNCT";
3042 tipc_printf(&pb, "Link <%s>\n"
3043 " %s MTU:%u Priority:%u Tolerance:%u ms"
3044 " Window:%u packets\n",
3045 l_ptr->name, status, link_max_pkt(l_ptr),
3046 l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3047 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3048 l_ptr->next_in_no - l_ptr->stats.recv_info,
3049 l_ptr->stats.recv_fragments,
3050 l_ptr->stats.recv_fragmented,
3051 l_ptr->stats.recv_bundles,
3052 l_ptr->stats.recv_bundled);
3053 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3054 l_ptr->next_out_no - l_ptr->stats.sent_info,
3055 l_ptr->stats.sent_fragments,
3056 l_ptr->stats.sent_fragmented,
3057 l_ptr->stats.sent_bundles,
3058 l_ptr->stats.sent_bundled);
3059 profile_total = l_ptr->stats.msg_length_counts;
3060 if (!profile_total)
3061 profile_total = 1;
3062 tipc_printf(&pb, " TX profile sample:%u packets average:%u octets\n"
3063 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3064 "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3065 l_ptr->stats.msg_length_counts,
3066 l_ptr->stats.msg_lengths_total / profile_total,
3067 percent(l_ptr->stats.msg_length_profile[0], profile_total),
3068 percent(l_ptr->stats.msg_length_profile[1], profile_total),
3069 percent(l_ptr->stats.msg_length_profile[2], profile_total),
3070 percent(l_ptr->stats.msg_length_profile[3], profile_total),
3071 percent(l_ptr->stats.msg_length_profile[4], profile_total),
3072 percent(l_ptr->stats.msg_length_profile[5], profile_total),
3073 percent(l_ptr->stats.msg_length_profile[6], profile_total));
3074 tipc_printf(&pb, " RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3075 l_ptr->stats.recv_states,
3076 l_ptr->stats.recv_probes,
3077 l_ptr->stats.recv_nacks,
3078 l_ptr->stats.deferred_recv,
3079 l_ptr->stats.duplicates);
3080 tipc_printf(&pb, " TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3081 l_ptr->stats.sent_states,
3082 l_ptr->stats.sent_probes,
3083 l_ptr->stats.sent_nacks,
3084 l_ptr->stats.sent_acks,
3085 l_ptr->stats.retransmitted);
3086 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
3087 l_ptr->stats.bearer_congs,
3088 l_ptr->stats.link_congs,
3089 l_ptr->stats.max_queue_sz,
3090 l_ptr->stats.queue_sz_counts
3091 ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3092 : 0);
3094 tipc_node_unlock(node);
3095 read_unlock_bh(&tipc_net_lock);
3096 return tipc_printbuf_validate(&pb);
3099 #define MAX_LINK_STATS_INFO 2000
3101 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3103 struct sk_buff *buf;
3104 struct tlv_desc *rep_tlv;
3105 int str_len;
3107 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3108 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3110 buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3111 if (!buf)
3112 return NULL;
3114 rep_tlv = (struct tlv_desc *)buf->data;
3116 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3117 (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3118 if (!str_len) {
3119 buf_discard(buf);
3120 return tipc_cfg_reply_error_string("link not found");
3123 skb_put(buf, TLV_SPACE(str_len));
3124 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3126 return buf;
3129 #if 0
3130 int link_control(const char *name, u32 op, u32 val)
3132 int res = -EINVAL;
3133 struct link *l_ptr;
3134 u32 bearer_id;
3135 struct node * node;
3136 u32 a;
3138 a = link_name2addr(name, &bearer_id);
3139 read_lock_bh(&tipc_net_lock);
3140 node = tipc_node_find(a);
3141 if (node) {
3142 tipc_node_lock(node);
3143 l_ptr = node->links[bearer_id];
3144 if (l_ptr) {
3145 if (op == TIPC_REMOVE_LINK) {
3146 struct bearer *b_ptr = l_ptr->b_ptr;
3147 spin_lock_bh(&b_ptr->publ.lock);
3148 tipc_link_delete(l_ptr);
3149 spin_unlock_bh(&b_ptr->publ.lock);
3151 if (op == TIPC_CMD_BLOCK_LINK) {
3152 tipc_link_reset(l_ptr);
3153 l_ptr->blocked = 1;
3155 if (op == TIPC_CMD_UNBLOCK_LINK) {
3156 l_ptr->blocked = 0;
3158 res = TIPC_OK;
3160 tipc_node_unlock(node);
3162 read_unlock_bh(&tipc_net_lock);
3163 return res;
3165 #endif
3168 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3169 * @dest: network address of destination node
3170 * @selector: used to select from set of active links
3172 * If no active link can be found, uses default maximum packet size.
3175 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3177 struct node *n_ptr;
3178 struct link *l_ptr;
3179 u32 res = MAX_PKT_DEFAULT;
3181 if (dest == tipc_own_addr)
3182 return MAX_MSG_SIZE;
3184 read_lock_bh(&tipc_net_lock);
3185 n_ptr = tipc_node_select(dest, selector);
3186 if (n_ptr) {
3187 tipc_node_lock(n_ptr);
3188 l_ptr = n_ptr->active_links[selector & 1];
3189 if (l_ptr)
3190 res = link_max_pkt(l_ptr);
3191 tipc_node_unlock(n_ptr);
3193 read_unlock_bh(&tipc_net_lock);
3194 return res;
3197 #if 0
3198 static void link_dump_rec_queue(struct link *l_ptr)
3200 struct sk_buff *crs;
3202 if (!l_ptr->oldest_deferred_in) {
3203 info("Reception queue empty\n");
3204 return;
3206 info("Contents of Reception queue:\n");
3207 crs = l_ptr->oldest_deferred_in;
3208 while (crs) {
3209 if (crs->data == (void *)0x0000a3a3) {
3210 info("buffer %x invalid\n", crs);
3211 return;
3213 msg_dbg(buf_msg(crs), "In rec queue: \n");
3214 crs = crs->next;
3217 #endif
3219 static void link_dump_send_queue(struct link *l_ptr)
3221 if (l_ptr->next_out) {
3222 info("\nContents of unsent queue:\n");
3223 dbg_print_buf_chain(l_ptr->next_out);
3225 info("\nContents of send queue:\n");
3226 if (l_ptr->first_out) {
3227 dbg_print_buf_chain(l_ptr->first_out);
3229 info("Empty send queue\n");
3232 static void link_print(struct link *l_ptr, struct print_buf *buf,
3233 const char *str)
3235 tipc_printf(buf, str);
3236 if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3237 return;
3238 tipc_printf(buf, "Link %x<%s>:",
3239 l_ptr->addr, l_ptr->b_ptr->publ.name);
3240 tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3241 tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3242 tipc_printf(buf, "SQUE");
3243 if (l_ptr->first_out) {
3244 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3245 if (l_ptr->next_out)
3246 tipc_printf(buf, "%u..",
3247 msg_seqno(buf_msg(l_ptr->next_out)));
3248 tipc_printf(buf, "%u]",
3249 msg_seqno(buf_msg
3250 (l_ptr->last_out)), l_ptr->out_queue_size);
3251 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3252 msg_seqno(buf_msg(l_ptr->first_out)))
3253 != (l_ptr->out_queue_size - 1))
3254 || (l_ptr->last_out->next != 0)) {
3255 tipc_printf(buf, "\nSend queue inconsistency\n");
3256 tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3257 tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3258 tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3259 link_dump_send_queue(l_ptr);
3261 } else
3262 tipc_printf(buf, "[]");
3263 tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3264 if (l_ptr->oldest_deferred_in) {
3265 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3266 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3267 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3268 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3269 tipc_printf(buf, ":RQSIZ(%u)",
3270 l_ptr->deferred_inqueue_sz);
3273 if (link_working_unknown(l_ptr))
3274 tipc_printf(buf, ":WU");
3275 if (link_reset_reset(l_ptr))
3276 tipc_printf(buf, ":RR");
3277 if (link_reset_unknown(l_ptr))
3278 tipc_printf(buf, ":RU");
3279 if (link_working_working(l_ptr))
3280 tipc_printf(buf, ":WW");
3281 tipc_printf(buf, "\n");