[ARM] 3704/1: format IOP Kconfig with tabs, create more consistency
[linux-2.6/kvm.git] / net / tipc / link.c
blobd64658053746f093d4429cdcd076666bf4bf99ce
1 /*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2006, Ericsson AB
5 * Copyright (c) 2004-2005, 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"
53 /*
54 * Limit for deferred reception queue:
57 #define DEF_QUEUE_LIMIT 256u
59 /*
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 */
67 /*
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
75 /*
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 */
100 /**
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 (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 1000 (bytes)
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)
163 link_print(l_ptr, DBG_OUTPUT, str);
166 static void dbg_print_buf_chain(struct sk_buff *root_buf)
168 if (DBG_OUTPUT) {
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 = (struct link *)kmalloc(sizeof(*l_ptr), GFP_ATOMIC);
421 if (!l_ptr) {
422 warn("Link creation failed, no memory\n");
423 return NULL;
425 memset(l_ptr, 0, sizeof(*l_ptr));
427 l_ptr->addr = peer;
428 if_name = strchr(b_ptr->publ.name, ':') + 1;
429 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
430 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
431 tipc_node(tipc_own_addr),
432 if_name,
433 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
434 /* note: peer i/f is appended to link name by reset/activate */
435 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
436 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
437 list_add_tail(&l_ptr->link_list, &b_ptr->links);
438 l_ptr->checkpoint = 1;
439 l_ptr->b_ptr = b_ptr;
440 link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
441 l_ptr->state = RESET_UNKNOWN;
443 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
444 msg = l_ptr->pmsg;
445 msg_init(msg, LINK_PROTOCOL, RESET_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
446 msg_set_size(msg, sizeof(l_ptr->proto_msg));
447 msg_set_session(msg, tipc_random);
448 msg_set_bearer_id(msg, b_ptr->identity);
449 strcpy((char *)msg_data(msg), if_name);
451 l_ptr->priority = b_ptr->priority;
452 tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
454 link_init_max_pkt(l_ptr);
456 l_ptr->next_out_no = 1;
457 INIT_LIST_HEAD(&l_ptr->waiting_ports);
459 link_reset_statistics(l_ptr);
461 l_ptr->owner = tipc_node_attach_link(l_ptr);
462 if (!l_ptr->owner) {
463 kfree(l_ptr);
464 return NULL;
467 if (LINK_LOG_BUF_SIZE) {
468 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
470 if (!pb) {
471 kfree(l_ptr);
472 warn("Link creation failed, no memory for print buffer\n");
473 return NULL;
475 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
478 tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
480 dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
481 l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
483 return l_ptr;
486 /**
487 * tipc_link_delete - delete a link
488 * @l_ptr: pointer to link
490 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
491 * This routine must not grab the node lock until after link timer cancellation
492 * to avoid a potential deadlock situation.
495 void tipc_link_delete(struct link *l_ptr)
497 if (!l_ptr) {
498 err("Attempt to delete non-existent link\n");
499 return;
502 dbg("tipc_link_delete()\n");
504 k_cancel_timer(&l_ptr->timer);
506 tipc_node_lock(l_ptr->owner);
507 tipc_link_reset(l_ptr);
508 tipc_node_detach_link(l_ptr->owner, l_ptr);
509 tipc_link_stop(l_ptr);
510 list_del_init(&l_ptr->link_list);
511 if (LINK_LOG_BUF_SIZE)
512 kfree(l_ptr->print_buf.buf);
513 tipc_node_unlock(l_ptr->owner);
514 k_term_timer(&l_ptr->timer);
515 kfree(l_ptr);
518 void tipc_link_start(struct link *l_ptr)
520 dbg("tipc_link_start %x\n", l_ptr);
521 link_state_event(l_ptr, STARTING_EVT);
525 * link_schedule_port - schedule port for deferred sending
526 * @l_ptr: pointer to link
527 * @origport: reference to sending port
528 * @sz: amount of data to be sent
530 * Schedules port for renewed sending of messages after link congestion
531 * has abated.
534 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
536 struct port *p_ptr;
538 spin_lock_bh(&tipc_port_list_lock);
539 p_ptr = tipc_port_lock(origport);
540 if (p_ptr) {
541 if (!p_ptr->wakeup)
542 goto exit;
543 if (!list_empty(&p_ptr->wait_list))
544 goto exit;
545 p_ptr->congested_link = l_ptr;
546 p_ptr->publ.congested = 1;
547 p_ptr->waiting_pkts = 1 + ((sz - 1) / link_max_pkt(l_ptr));
548 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
549 l_ptr->stats.link_congs++;
550 exit:
551 tipc_port_unlock(p_ptr);
553 spin_unlock_bh(&tipc_port_list_lock);
554 return -ELINKCONG;
557 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
559 struct port *p_ptr;
560 struct port *temp_p_ptr;
561 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
563 if (all)
564 win = 100000;
565 if (win <= 0)
566 return;
567 if (!spin_trylock_bh(&tipc_port_list_lock))
568 return;
569 if (link_congested(l_ptr))
570 goto exit;
571 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
572 wait_list) {
573 if (win <= 0)
574 break;
575 list_del_init(&p_ptr->wait_list);
576 p_ptr->congested_link = NULL;
577 spin_lock_bh(p_ptr->publ.lock);
578 p_ptr->publ.congested = 0;
579 p_ptr->wakeup(&p_ptr->publ);
580 win -= p_ptr->waiting_pkts;
581 spin_unlock_bh(p_ptr->publ.lock);
584 exit:
585 spin_unlock_bh(&tipc_port_list_lock);
588 /**
589 * link_release_outqueue - purge link's outbound message queue
590 * @l_ptr: pointer to link
593 static void link_release_outqueue(struct link *l_ptr)
595 struct sk_buff *buf = l_ptr->first_out;
596 struct sk_buff *next;
598 while (buf) {
599 next = buf->next;
600 buf_discard(buf);
601 buf = next;
603 l_ptr->first_out = NULL;
604 l_ptr->out_queue_size = 0;
608 * tipc_link_reset_fragments - purge link's inbound message fragments queue
609 * @l_ptr: pointer to link
612 void tipc_link_reset_fragments(struct link *l_ptr)
614 struct sk_buff *buf = l_ptr->defragm_buf;
615 struct sk_buff *next;
617 while (buf) {
618 next = buf->next;
619 buf_discard(buf);
620 buf = next;
622 l_ptr->defragm_buf = NULL;
625 /**
626 * tipc_link_stop - purge all inbound and outbound messages associated with link
627 * @l_ptr: pointer to link
630 void tipc_link_stop(struct link *l_ptr)
632 struct sk_buff *buf;
633 struct sk_buff *next;
635 buf = l_ptr->oldest_deferred_in;
636 while (buf) {
637 next = buf->next;
638 buf_discard(buf);
639 buf = next;
642 buf = l_ptr->first_out;
643 while (buf) {
644 next = buf->next;
645 buf_discard(buf);
646 buf = next;
649 tipc_link_reset_fragments(l_ptr);
651 buf_discard(l_ptr->proto_msg_queue);
652 l_ptr->proto_msg_queue = NULL;
655 #if 0
657 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
659 static void link_recv_event(struct link_event *ev)
661 ev->fcn(ev->addr, ev->name, ev->up);
662 kfree(ev);
665 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
666 struct link *l_ptr, int up)
668 struct link_event *ev;
670 ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
671 if (!ev) {
672 warn("Link event allocation failure\n");
673 return;
675 ev->addr = l_ptr->addr;
676 ev->up = up;
677 ev->fcn = fcn;
678 memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
679 tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
682 #else
684 #define link_send_event(fcn, l_ptr, up) do { } while (0)
686 #endif
688 void tipc_link_reset(struct link *l_ptr)
690 struct sk_buff *buf;
691 u32 prev_state = l_ptr->state;
692 u32 checkpoint = l_ptr->next_in_no;
693 int was_active_link = tipc_link_is_active(l_ptr);
695 msg_set_session(l_ptr->pmsg, msg_session(l_ptr->pmsg) + 1);
697 /* Link is down, accept any session: */
698 l_ptr->peer_session = 0;
700 /* Prepare for max packet size negotiation */
701 link_init_max_pkt(l_ptr);
703 l_ptr->state = RESET_UNKNOWN;
704 dbg_link_state("Resetting Link\n");
706 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
707 return;
709 tipc_node_link_down(l_ptr->owner, l_ptr);
710 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
711 #if 0
712 tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
713 dbg_link_dump();
714 #endif
715 if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
716 l_ptr->owner->permit_changeover) {
717 l_ptr->reset_checkpoint = checkpoint;
718 l_ptr->exp_msg_count = START_CHANGEOVER;
721 /* Clean up all queues: */
723 link_release_outqueue(l_ptr);
724 buf_discard(l_ptr->proto_msg_queue);
725 l_ptr->proto_msg_queue = NULL;
726 buf = l_ptr->oldest_deferred_in;
727 while (buf) {
728 struct sk_buff *next = buf->next;
729 buf_discard(buf);
730 buf = next;
732 if (!list_empty(&l_ptr->waiting_ports))
733 tipc_link_wakeup_ports(l_ptr, 1);
735 l_ptr->retransm_queue_head = 0;
736 l_ptr->retransm_queue_size = 0;
737 l_ptr->last_out = NULL;
738 l_ptr->first_out = NULL;
739 l_ptr->next_out = NULL;
740 l_ptr->unacked_window = 0;
741 l_ptr->checkpoint = 1;
742 l_ptr->next_out_no = 1;
743 l_ptr->deferred_inqueue_sz = 0;
744 l_ptr->oldest_deferred_in = NULL;
745 l_ptr->newest_deferred_in = NULL;
746 l_ptr->fsm_msg_cnt = 0;
747 l_ptr->stale_count = 0;
748 link_reset_statistics(l_ptr);
750 link_send_event(tipc_cfg_link_event, l_ptr, 0);
751 if (!in_own_cluster(l_ptr->addr))
752 link_send_event(tipc_disc_link_event, l_ptr, 0);
756 static void link_activate(struct link *l_ptr)
758 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
759 tipc_node_link_up(l_ptr->owner, l_ptr);
760 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
761 link_send_event(tipc_cfg_link_event, l_ptr, 1);
762 if (!in_own_cluster(l_ptr->addr))
763 link_send_event(tipc_disc_link_event, l_ptr, 1);
767 * link_state_event - link finite state machine
768 * @l_ptr: pointer to link
769 * @event: state machine event to process
772 static void link_state_event(struct link *l_ptr, unsigned event)
774 struct link *other;
775 u32 cont_intv = l_ptr->continuity_interval;
777 if (!l_ptr->started && (event != STARTING_EVT))
778 return; /* Not yet. */
780 if (link_blocked(l_ptr)) {
781 if (event == TIMEOUT_EVT) {
782 link_set_timer(l_ptr, cont_intv);
784 return; /* Changeover going on */
786 dbg_link("STATE_EV: <%s> ", l_ptr->name);
788 switch (l_ptr->state) {
789 case WORKING_WORKING:
790 dbg_link("WW/");
791 switch (event) {
792 case TRAFFIC_MSG_EVT:
793 dbg_link("TRF-");
794 /* fall through */
795 case ACTIVATE_MSG:
796 dbg_link("ACT\n");
797 break;
798 case TIMEOUT_EVT:
799 dbg_link("TIM ");
800 if (l_ptr->next_in_no != l_ptr->checkpoint) {
801 l_ptr->checkpoint = l_ptr->next_in_no;
802 if (tipc_bclink_acks_missing(l_ptr->owner)) {
803 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
804 0, 0, 0, 0, 0);
805 l_ptr->fsm_msg_cnt++;
806 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
807 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
808 1, 0, 0, 0, 0);
809 l_ptr->fsm_msg_cnt++;
811 link_set_timer(l_ptr, cont_intv);
812 break;
814 dbg_link(" -> WU\n");
815 l_ptr->state = WORKING_UNKNOWN;
816 l_ptr->fsm_msg_cnt = 0;
817 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
818 l_ptr->fsm_msg_cnt++;
819 link_set_timer(l_ptr, cont_intv / 4);
820 break;
821 case RESET_MSG:
822 dbg_link("RES -> RR\n");
823 info("Resetting link <%s>, requested by peer\n",
824 l_ptr->name);
825 tipc_link_reset(l_ptr);
826 l_ptr->state = RESET_RESET;
827 l_ptr->fsm_msg_cnt = 0;
828 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
829 l_ptr->fsm_msg_cnt++;
830 link_set_timer(l_ptr, cont_intv);
831 break;
832 default:
833 err("Unknown link event %u in WW state\n", event);
835 break;
836 case WORKING_UNKNOWN:
837 dbg_link("WU/");
838 switch (event) {
839 case TRAFFIC_MSG_EVT:
840 dbg_link("TRF-");
841 case ACTIVATE_MSG:
842 dbg_link("ACT -> WW\n");
843 l_ptr->state = WORKING_WORKING;
844 l_ptr->fsm_msg_cnt = 0;
845 link_set_timer(l_ptr, cont_intv);
846 break;
847 case RESET_MSG:
848 dbg_link("RES -> RR\n");
849 info("Resetting link <%s>, requested by peer "
850 "while probing\n", l_ptr->name);
851 tipc_link_reset(l_ptr);
852 l_ptr->state = RESET_RESET;
853 l_ptr->fsm_msg_cnt = 0;
854 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
855 l_ptr->fsm_msg_cnt++;
856 link_set_timer(l_ptr, cont_intv);
857 break;
858 case TIMEOUT_EVT:
859 dbg_link("TIM ");
860 if (l_ptr->next_in_no != l_ptr->checkpoint) {
861 dbg_link("-> WW \n");
862 l_ptr->state = WORKING_WORKING;
863 l_ptr->fsm_msg_cnt = 0;
864 l_ptr->checkpoint = l_ptr->next_in_no;
865 if (tipc_bclink_acks_missing(l_ptr->owner)) {
866 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
867 0, 0, 0, 0, 0);
868 l_ptr->fsm_msg_cnt++;
870 link_set_timer(l_ptr, cont_intv);
871 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
872 dbg_link("Probing %u/%u,timer = %u ms)\n",
873 l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
874 cont_intv / 4);
875 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
876 1, 0, 0, 0, 0);
877 l_ptr->fsm_msg_cnt++;
878 link_set_timer(l_ptr, cont_intv / 4);
879 } else { /* Link has failed */
880 dbg_link("-> RU (%u probes unanswered)\n",
881 l_ptr->fsm_msg_cnt);
882 warn("Resetting link <%s>, peer not responding\n",
883 l_ptr->name);
884 tipc_link_reset(l_ptr);
885 l_ptr->state = RESET_UNKNOWN;
886 l_ptr->fsm_msg_cnt = 0;
887 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
888 0, 0, 0, 0, 0);
889 l_ptr->fsm_msg_cnt++;
890 link_set_timer(l_ptr, cont_intv);
892 break;
893 default:
894 err("Unknown link event %u in WU state\n", event);
896 break;
897 case RESET_UNKNOWN:
898 dbg_link("RU/");
899 switch (event) {
900 case TRAFFIC_MSG_EVT:
901 dbg_link("TRF-\n");
902 break;
903 case ACTIVATE_MSG:
904 other = l_ptr->owner->active_links[0];
905 if (other && link_working_unknown(other)) {
906 dbg_link("ACT\n");
907 break;
909 dbg_link("ACT -> WW\n");
910 l_ptr->state = WORKING_WORKING;
911 l_ptr->fsm_msg_cnt = 0;
912 link_activate(l_ptr);
913 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
914 l_ptr->fsm_msg_cnt++;
915 link_set_timer(l_ptr, cont_intv);
916 break;
917 case RESET_MSG:
918 dbg_link("RES \n");
919 dbg_link(" -> RR\n");
920 l_ptr->state = RESET_RESET;
921 l_ptr->fsm_msg_cnt = 0;
922 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
923 l_ptr->fsm_msg_cnt++;
924 link_set_timer(l_ptr, cont_intv);
925 break;
926 case STARTING_EVT:
927 dbg_link("START-");
928 l_ptr->started = 1;
929 /* fall through */
930 case TIMEOUT_EVT:
931 dbg_link("TIM \n");
932 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
933 l_ptr->fsm_msg_cnt++;
934 link_set_timer(l_ptr, cont_intv);
935 break;
936 default:
937 err("Unknown link event %u in RU state\n", event);
939 break;
940 case RESET_RESET:
941 dbg_link("RR/ ");
942 switch (event) {
943 case TRAFFIC_MSG_EVT:
944 dbg_link("TRF-");
945 /* fall through */
946 case ACTIVATE_MSG:
947 other = l_ptr->owner->active_links[0];
948 if (other && link_working_unknown(other)) {
949 dbg_link("ACT\n");
950 break;
952 dbg_link("ACT -> WW\n");
953 l_ptr->state = WORKING_WORKING;
954 l_ptr->fsm_msg_cnt = 0;
955 link_activate(l_ptr);
956 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
957 l_ptr->fsm_msg_cnt++;
958 link_set_timer(l_ptr, cont_intv);
959 break;
960 case RESET_MSG:
961 dbg_link("RES\n");
962 break;
963 case TIMEOUT_EVT:
964 dbg_link("TIM\n");
965 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
966 l_ptr->fsm_msg_cnt++;
967 link_set_timer(l_ptr, cont_intv);
968 dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
969 break;
970 default:
971 err("Unknown link event %u in RR state\n", event);
973 break;
974 default:
975 err("Unknown link state %u/%u\n", l_ptr->state, event);
980 * link_bundle_buf(): Append contents of a buffer to
981 * the tail of an existing one.
984 static int link_bundle_buf(struct link *l_ptr,
985 struct sk_buff *bundler,
986 struct sk_buff *buf)
988 struct tipc_msg *bundler_msg = buf_msg(bundler);
989 struct tipc_msg *msg = buf_msg(buf);
990 u32 size = msg_size(msg);
991 u32 to_pos = align(msg_size(bundler_msg));
992 u32 rest = link_max_pkt(l_ptr) - to_pos;
994 if (msg_user(bundler_msg) != MSG_BUNDLER)
995 return 0;
996 if (msg_type(bundler_msg) != OPEN_MSG)
997 return 0;
998 if (rest < align(size))
999 return 0;
1001 skb_put(bundler, (to_pos - msg_size(bundler_msg)) + size);
1002 memcpy(bundler->data + to_pos, buf->data, size);
1003 msg_set_size(bundler_msg, to_pos + size);
1004 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
1005 dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
1006 msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
1007 msg_dbg(msg, "PACKD:");
1008 buf_discard(buf);
1009 l_ptr->stats.sent_bundled++;
1010 return 1;
1013 static void link_add_to_outqueue(struct link *l_ptr,
1014 struct sk_buff *buf,
1015 struct tipc_msg *msg)
1017 u32 ack = mod(l_ptr->next_in_no - 1);
1018 u32 seqno = mod(l_ptr->next_out_no++);
1020 msg_set_word(msg, 2, ((ack << 16) | seqno));
1021 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1022 buf->next = NULL;
1023 if (l_ptr->first_out) {
1024 l_ptr->last_out->next = buf;
1025 l_ptr->last_out = buf;
1026 } else
1027 l_ptr->first_out = l_ptr->last_out = buf;
1028 l_ptr->out_queue_size++;
1032 * tipc_link_send_buf() is the 'full path' for messages, called from
1033 * inside TIPC when the 'fast path' in tipc_send_buf
1034 * has failed, and from link_send()
1037 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1039 struct tipc_msg *msg = buf_msg(buf);
1040 u32 size = msg_size(msg);
1041 u32 dsz = msg_data_sz(msg);
1042 u32 queue_size = l_ptr->out_queue_size;
1043 u32 imp = msg_tot_importance(msg);
1044 u32 queue_limit = l_ptr->queue_limit[imp];
1045 u32 max_packet = link_max_pkt(l_ptr);
1047 msg_set_prevnode(msg, tipc_own_addr); /* If routed message */
1049 /* Match msg importance against queue limits: */
1051 if (unlikely(queue_size >= queue_limit)) {
1052 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1053 return link_schedule_port(l_ptr, msg_origport(msg),
1054 size);
1056 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1057 buf_discard(buf);
1058 if (imp > CONN_MANAGER) {
1059 warn("Resetting link <%s>, send queue full", l_ptr->name);
1060 tipc_link_reset(l_ptr);
1062 return dsz;
1065 /* Fragmentation needed ? */
1067 if (size > max_packet)
1068 return tipc_link_send_long_buf(l_ptr, buf);
1070 /* Packet can be queued or sent: */
1072 if (queue_size > l_ptr->stats.max_queue_sz)
1073 l_ptr->stats.max_queue_sz = queue_size;
1075 if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1076 !link_congested(l_ptr))) {
1077 link_add_to_outqueue(l_ptr, buf, msg);
1079 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1080 l_ptr->unacked_window = 0;
1081 } else {
1082 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1083 l_ptr->stats.bearer_congs++;
1084 l_ptr->next_out = buf;
1086 return dsz;
1088 /* Congestion: can message be bundled ?: */
1090 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1091 (msg_user(msg) != MSG_FRAGMENTER)) {
1093 /* Try adding message to an existing bundle */
1095 if (l_ptr->next_out &&
1096 link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1097 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1098 return dsz;
1101 /* Try creating a new bundle */
1103 if (size <= max_packet * 2 / 3) {
1104 struct sk_buff *bundler = buf_acquire(max_packet);
1105 struct tipc_msg bundler_hdr;
1107 if (bundler) {
1108 msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1109 TIPC_OK, INT_H_SIZE, l_ptr->addr);
1110 memcpy(bundler->data, (unchar *)&bundler_hdr,
1111 INT_H_SIZE);
1112 skb_trim(bundler, INT_H_SIZE);
1113 link_bundle_buf(l_ptr, bundler, buf);
1114 buf = bundler;
1115 msg = buf_msg(buf);
1116 l_ptr->stats.sent_bundles++;
1120 if (!l_ptr->next_out)
1121 l_ptr->next_out = buf;
1122 link_add_to_outqueue(l_ptr, buf, msg);
1123 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1124 return dsz;
1128 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1129 * not been selected yet, and the the owner node is not locked
1130 * Called by TIPC internal users, e.g. the name distributor
1133 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1135 struct link *l_ptr;
1136 struct node *n_ptr;
1137 int res = -ELINKCONG;
1139 read_lock_bh(&tipc_net_lock);
1140 n_ptr = tipc_node_select(dest, selector);
1141 if (n_ptr) {
1142 tipc_node_lock(n_ptr);
1143 l_ptr = n_ptr->active_links[selector & 1];
1144 if (l_ptr) {
1145 dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1146 res = tipc_link_send_buf(l_ptr, buf);
1147 } else {
1148 dbg("Attempt to send msg to unreachable node:\n");
1149 msg_dbg(buf_msg(buf),">>>");
1150 buf_discard(buf);
1152 tipc_node_unlock(n_ptr);
1153 } else {
1154 dbg("Attempt to send msg to unknown node:\n");
1155 msg_dbg(buf_msg(buf),">>>");
1156 buf_discard(buf);
1158 read_unlock_bh(&tipc_net_lock);
1159 return res;
1163 * link_send_buf_fast: Entry for data messages where the
1164 * destination link is known and the header is complete,
1165 * inclusive total message length. Very time critical.
1166 * Link is locked. Returns user data length.
1169 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1170 u32 *used_max_pkt)
1172 struct tipc_msg *msg = buf_msg(buf);
1173 int res = msg_data_sz(msg);
1175 if (likely(!link_congested(l_ptr))) {
1176 if (likely(msg_size(msg) <= link_max_pkt(l_ptr))) {
1177 if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1178 link_add_to_outqueue(l_ptr, buf, msg);
1179 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1180 &l_ptr->media_addr))) {
1181 l_ptr->unacked_window = 0;
1182 msg_dbg(msg,"SENT_FAST:");
1183 return res;
1185 dbg("failed sent fast...\n");
1186 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1187 l_ptr->stats.bearer_congs++;
1188 l_ptr->next_out = buf;
1189 return res;
1192 else
1193 *used_max_pkt = link_max_pkt(l_ptr);
1195 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
1199 * tipc_send_buf_fast: Entry for data messages where the
1200 * destination node is known and the header is complete,
1201 * inclusive total message length.
1202 * Returns user data length.
1204 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1206 struct link *l_ptr;
1207 struct node *n_ptr;
1208 int res;
1209 u32 selector = msg_origport(buf_msg(buf)) & 1;
1210 u32 dummy;
1212 if (destnode == tipc_own_addr)
1213 return tipc_port_recv_msg(buf);
1215 read_lock_bh(&tipc_net_lock);
1216 n_ptr = tipc_node_select(destnode, selector);
1217 if (likely(n_ptr)) {
1218 tipc_node_lock(n_ptr);
1219 l_ptr = n_ptr->active_links[selector];
1220 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1221 buf, l_ptr, destnode);
1222 if (likely(l_ptr)) {
1223 res = link_send_buf_fast(l_ptr, buf, &dummy);
1224 tipc_node_unlock(n_ptr);
1225 read_unlock_bh(&tipc_net_lock);
1226 return res;
1228 tipc_node_unlock(n_ptr);
1230 read_unlock_bh(&tipc_net_lock);
1231 res = msg_data_sz(buf_msg(buf));
1232 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1233 return res;
1238 * tipc_link_send_sections_fast: Entry for messages where the
1239 * destination processor is known and the header is complete,
1240 * except for total message length.
1241 * Returns user data length or errno.
1243 int tipc_link_send_sections_fast(struct port *sender,
1244 struct iovec const *msg_sect,
1245 const u32 num_sect,
1246 u32 destaddr)
1248 struct tipc_msg *hdr = &sender->publ.phdr;
1249 struct link *l_ptr;
1250 struct sk_buff *buf;
1251 struct node *node;
1252 int res;
1253 u32 selector = msg_origport(hdr) & 1;
1255 again:
1257 * Try building message using port's max_pkt hint.
1258 * (Must not hold any locks while building message.)
1261 res = msg_build(hdr, msg_sect, num_sect, sender->max_pkt,
1262 !sender->user_port, &buf);
1264 read_lock_bh(&tipc_net_lock);
1265 node = tipc_node_select(destaddr, selector);
1266 if (likely(node)) {
1267 tipc_node_lock(node);
1268 l_ptr = node->active_links[selector];
1269 if (likely(l_ptr)) {
1270 if (likely(buf)) {
1271 res = link_send_buf_fast(l_ptr, buf,
1272 &sender->max_pkt);
1273 if (unlikely(res < 0))
1274 buf_discard(buf);
1275 exit:
1276 tipc_node_unlock(node);
1277 read_unlock_bh(&tipc_net_lock);
1278 return res;
1281 /* Exit if build request was invalid */
1283 if (unlikely(res < 0))
1284 goto exit;
1286 /* Exit if link (or bearer) is congested */
1288 if (link_congested(l_ptr) ||
1289 !list_empty(&l_ptr->b_ptr->cong_links)) {
1290 res = link_schedule_port(l_ptr,
1291 sender->publ.ref, res);
1292 goto exit;
1296 * Message size exceeds max_pkt hint; update hint,
1297 * then re-try fast path or fragment the message
1300 sender->max_pkt = link_max_pkt(l_ptr);
1301 tipc_node_unlock(node);
1302 read_unlock_bh(&tipc_net_lock);
1305 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1306 goto again;
1308 return link_send_sections_long(sender, msg_sect,
1309 num_sect, destaddr);
1311 tipc_node_unlock(node);
1313 read_unlock_bh(&tipc_net_lock);
1315 /* Couldn't find a link to the destination node */
1317 if (buf)
1318 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1319 if (res >= 0)
1320 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1321 TIPC_ERR_NO_NODE);
1322 return res;
1326 * link_send_sections_long(): Entry for long messages where the
1327 * destination node is known and the header is complete,
1328 * inclusive total message length.
1329 * Link and bearer congestion status have been checked to be ok,
1330 * and are ignored if they change.
1332 * Note that fragments do not use the full link MTU so that they won't have
1333 * to undergo refragmentation if link changeover causes them to be sent
1334 * over another link with an additional tunnel header added as prefix.
1335 * (Refragmentation will still occur if the other link has a smaller MTU.)
1337 * Returns user data length or errno.
1339 static int link_send_sections_long(struct port *sender,
1340 struct iovec const *msg_sect,
1341 u32 num_sect,
1342 u32 destaddr)
1344 struct link *l_ptr;
1345 struct node *node;
1346 struct tipc_msg *hdr = &sender->publ.phdr;
1347 u32 dsz = msg_data_sz(hdr);
1348 u32 max_pkt,fragm_sz,rest;
1349 struct tipc_msg fragm_hdr;
1350 struct sk_buff *buf,*buf_chain,*prev;
1351 u32 fragm_crs,fragm_rest,hsz,sect_rest;
1352 const unchar *sect_crs;
1353 int curr_sect;
1354 u32 fragm_no;
1356 again:
1357 fragm_no = 1;
1358 max_pkt = sender->max_pkt - INT_H_SIZE;
1359 /* leave room for tunnel header in case of link changeover */
1360 fragm_sz = max_pkt - INT_H_SIZE;
1361 /* leave room for fragmentation header in each fragment */
1362 rest = dsz;
1363 fragm_crs = 0;
1364 fragm_rest = 0;
1365 sect_rest = 0;
1366 sect_crs = NULL;
1367 curr_sect = -1;
1369 /* Prepare reusable fragment header: */
1371 msg_dbg(hdr, ">FRAGMENTING>");
1372 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1373 TIPC_OK, INT_H_SIZE, msg_destnode(hdr));
1374 msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1375 msg_set_size(&fragm_hdr, max_pkt);
1376 msg_set_fragm_no(&fragm_hdr, 1);
1378 /* Prepare header of first fragment: */
1380 buf_chain = buf = buf_acquire(max_pkt);
1381 if (!buf)
1382 return -ENOMEM;
1383 buf->next = NULL;
1384 memcpy(buf->data, (unchar *)&fragm_hdr, INT_H_SIZE);
1385 hsz = msg_hdr_sz(hdr);
1386 memcpy(buf->data + INT_H_SIZE, (unchar *)hdr, hsz);
1387 msg_dbg(buf_msg(buf), ">BUILD>");
1389 /* Chop up message: */
1391 fragm_crs = INT_H_SIZE + hsz;
1392 fragm_rest = fragm_sz - hsz;
1394 do { /* For all sections */
1395 u32 sz;
1397 if (!sect_rest) {
1398 sect_rest = msg_sect[++curr_sect].iov_len;
1399 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1402 if (sect_rest < fragm_rest)
1403 sz = sect_rest;
1404 else
1405 sz = fragm_rest;
1407 if (likely(!sender->user_port)) {
1408 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1409 error:
1410 for (; buf_chain; buf_chain = buf) {
1411 buf = buf_chain->next;
1412 buf_discard(buf_chain);
1414 return -EFAULT;
1416 } else
1417 memcpy(buf->data + fragm_crs, sect_crs, sz);
1419 sect_crs += sz;
1420 sect_rest -= sz;
1421 fragm_crs += sz;
1422 fragm_rest -= sz;
1423 rest -= sz;
1425 if (!fragm_rest && rest) {
1427 /* Initiate new fragment: */
1428 if (rest <= fragm_sz) {
1429 fragm_sz = rest;
1430 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1431 } else {
1432 msg_set_type(&fragm_hdr, FRAGMENT);
1434 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1435 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1436 prev = buf;
1437 buf = buf_acquire(fragm_sz + INT_H_SIZE);
1438 if (!buf)
1439 goto error;
1441 buf->next = NULL;
1442 prev->next = buf;
1443 memcpy(buf->data, (unchar *)&fragm_hdr, INT_H_SIZE);
1444 fragm_crs = INT_H_SIZE;
1445 fragm_rest = fragm_sz;
1446 msg_dbg(buf_msg(buf)," >BUILD>");
1449 while (rest > 0);
1452 * Now we have a buffer chain. Select a link and check
1453 * that packet size is still OK
1455 node = tipc_node_select(destaddr, sender->publ.ref & 1);
1456 if (likely(node)) {
1457 tipc_node_lock(node);
1458 l_ptr = node->active_links[sender->publ.ref & 1];
1459 if (!l_ptr) {
1460 tipc_node_unlock(node);
1461 goto reject;
1463 if (link_max_pkt(l_ptr) < max_pkt) {
1464 sender->max_pkt = link_max_pkt(l_ptr);
1465 tipc_node_unlock(node);
1466 for (; buf_chain; buf_chain = buf) {
1467 buf = buf_chain->next;
1468 buf_discard(buf_chain);
1470 goto again;
1472 } else {
1473 reject:
1474 for (; buf_chain; buf_chain = buf) {
1475 buf = buf_chain->next;
1476 buf_discard(buf_chain);
1478 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1479 TIPC_ERR_NO_NODE);
1482 /* Append whole chain to send queue: */
1484 buf = buf_chain;
1485 l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1486 if (!l_ptr->next_out)
1487 l_ptr->next_out = buf_chain;
1488 l_ptr->stats.sent_fragmented++;
1489 while (buf) {
1490 struct sk_buff *next = buf->next;
1491 struct tipc_msg *msg = buf_msg(buf);
1493 l_ptr->stats.sent_fragments++;
1494 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1495 link_add_to_outqueue(l_ptr, buf, msg);
1496 msg_dbg(msg, ">ADD>");
1497 buf = next;
1500 /* Send it, if possible: */
1502 tipc_link_push_queue(l_ptr);
1503 tipc_node_unlock(node);
1504 return dsz;
1508 * tipc_link_push_packet: Push one unsent packet to the media
1510 u32 tipc_link_push_packet(struct link *l_ptr)
1512 struct sk_buff *buf = l_ptr->first_out;
1513 u32 r_q_size = l_ptr->retransm_queue_size;
1514 u32 r_q_head = l_ptr->retransm_queue_head;
1516 /* Step to position where retransmission failed, if any, */
1517 /* consider that buffers may have been released in meantime */
1519 if (r_q_size && buf) {
1520 u32 last = lesser(mod(r_q_head + r_q_size),
1521 link_last_sent(l_ptr));
1522 u32 first = msg_seqno(buf_msg(buf));
1524 while (buf && less(first, r_q_head)) {
1525 first = mod(first + 1);
1526 buf = buf->next;
1528 l_ptr->retransm_queue_head = r_q_head = first;
1529 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1532 /* Continue retransmission now, if there is anything: */
1534 if (r_q_size && buf && !skb_cloned(buf)) {
1535 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1536 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1537 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1538 msg_dbg(buf_msg(buf), ">DEF-RETR>");
1539 l_ptr->retransm_queue_head = mod(++r_q_head);
1540 l_ptr->retransm_queue_size = --r_q_size;
1541 l_ptr->stats.retransmitted++;
1542 return TIPC_OK;
1543 } else {
1544 l_ptr->stats.bearer_congs++;
1545 msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1546 return PUSH_FAILED;
1550 /* Send deferred protocol message, if any: */
1552 buf = l_ptr->proto_msg_queue;
1553 if (buf) {
1554 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1555 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1556 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1557 msg_dbg(buf_msg(buf), ">DEF-PROT>");
1558 l_ptr->unacked_window = 0;
1559 buf_discard(buf);
1560 l_ptr->proto_msg_queue = NULL;
1561 return TIPC_OK;
1562 } else {
1563 msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1564 l_ptr->stats.bearer_congs++;
1565 return PUSH_FAILED;
1569 /* Send one deferred data message, if send window not full: */
1571 buf = l_ptr->next_out;
1572 if (buf) {
1573 struct tipc_msg *msg = buf_msg(buf);
1574 u32 next = msg_seqno(msg);
1575 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1577 if (mod(next - first) < l_ptr->queue_limit[0]) {
1578 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1579 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1580 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1581 if (msg_user(msg) == MSG_BUNDLER)
1582 msg_set_type(msg, CLOSED_MSG);
1583 msg_dbg(msg, ">PUSH-DATA>");
1584 l_ptr->next_out = buf->next;
1585 return TIPC_OK;
1586 } else {
1587 msg_dbg(msg, "|PUSH-DATA|");
1588 l_ptr->stats.bearer_congs++;
1589 return PUSH_FAILED;
1593 return PUSH_FINISHED;
1597 * push_queue(): push out the unsent messages of a link where
1598 * congestion has abated. Node is locked
1600 void tipc_link_push_queue(struct link *l_ptr)
1602 u32 res;
1604 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1605 return;
1607 do {
1608 res = tipc_link_push_packet(l_ptr);
1610 while (res == TIPC_OK);
1611 if (res == PUSH_FAILED)
1612 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1615 static void link_reset_all(unsigned long addr)
1617 struct node *n_ptr;
1618 char addr_string[16];
1619 u32 i;
1621 read_lock_bh(&tipc_net_lock);
1622 n_ptr = tipc_node_find((u32)addr);
1623 if (!n_ptr) {
1624 read_unlock_bh(&tipc_net_lock);
1625 return; /* node no longer exists */
1628 tipc_node_lock(n_ptr);
1630 warn("Resetting all links to %s\n",
1631 addr_string_fill(addr_string, n_ptr->addr));
1633 for (i = 0; i < MAX_BEARERS; i++) {
1634 if (n_ptr->links[i]) {
1635 link_print(n_ptr->links[i], TIPC_OUTPUT,
1636 "Resetting link\n");
1637 tipc_link_reset(n_ptr->links[i]);
1641 tipc_node_unlock(n_ptr);
1642 read_unlock_bh(&tipc_net_lock);
1645 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1647 struct tipc_msg *msg = buf_msg(buf);
1649 warn("Retransmission failure on link <%s>\n", l_ptr->name);
1650 tipc_msg_print(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1652 if (l_ptr->addr) {
1654 /* Handle failure on standard link */
1656 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1657 tipc_link_reset(l_ptr);
1659 } else {
1661 /* Handle failure on broadcast link */
1663 struct node *n_ptr;
1664 char addr_string[16];
1666 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u, ", msg_seqno(msg));
1667 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %u\n", (u32)TIPC_SKB_CB(buf)->handle);
1669 n_ptr = l_ptr->owner->next;
1670 tipc_node_lock(n_ptr);
1672 addr_string_fill(addr_string, n_ptr->addr);
1673 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1674 tipc_printf(TIPC_OUTPUT, "Supported: %d, ", n_ptr->bclink.supported);
1675 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1676 tipc_printf(TIPC_OUTPUT, "Last in: %u, ", n_ptr->bclink.last_in);
1677 tipc_printf(TIPC_OUTPUT, "Gap after: %u, ", n_ptr->bclink.gap_after);
1678 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1679 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1681 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1683 tipc_node_unlock(n_ptr);
1685 l_ptr->stale_count = 0;
1689 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1690 u32 retransmits)
1692 struct tipc_msg *msg;
1694 if (!buf)
1695 return;
1697 msg = buf_msg(buf);
1699 dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1701 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1702 if (!skb_cloned(buf)) {
1703 msg_dbg(msg, ">NO_RETR->BCONG>");
1704 dbg_print_link(l_ptr, " ");
1705 l_ptr->retransm_queue_head = msg_seqno(msg);
1706 l_ptr->retransm_queue_size = retransmits;
1707 return;
1708 } else {
1709 /* Don't retransmit if driver already has the buffer */
1711 } else {
1712 /* Detect repeated retransmit failures on uncongested bearer */
1714 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1715 if (++l_ptr->stale_count > 100) {
1716 link_retransmit_failure(l_ptr, buf);
1717 return;
1719 } else {
1720 l_ptr->last_retransmitted = msg_seqno(msg);
1721 l_ptr->stale_count = 1;
1725 while (retransmits && (buf != l_ptr->next_out) && buf && !skb_cloned(buf)) {
1726 msg = buf_msg(buf);
1727 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1728 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1729 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1730 msg_dbg(buf_msg(buf), ">RETR>");
1731 buf = buf->next;
1732 retransmits--;
1733 l_ptr->stats.retransmitted++;
1734 } else {
1735 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1736 l_ptr->stats.bearer_congs++;
1737 l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1738 l_ptr->retransm_queue_size = retransmits;
1739 return;
1743 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1747 * link_recv_non_seq: Receive packets which are outside
1748 * the link sequence flow
1751 static void link_recv_non_seq(struct sk_buff *buf)
1753 struct tipc_msg *msg = buf_msg(buf);
1755 if (msg_user(msg) == LINK_CONFIG)
1756 tipc_disc_recv_msg(buf);
1757 else
1758 tipc_bclink_recv_pkt(buf);
1761 /**
1762 * link_insert_deferred_queue - insert deferred messages back into receive chain
1765 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1766 struct sk_buff *buf)
1768 u32 seq_no;
1770 if (l_ptr->oldest_deferred_in == NULL)
1771 return buf;
1773 seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1774 if (seq_no == mod(l_ptr->next_in_no)) {
1775 l_ptr->newest_deferred_in->next = buf;
1776 buf = l_ptr->oldest_deferred_in;
1777 l_ptr->oldest_deferred_in = NULL;
1778 l_ptr->deferred_inqueue_sz = 0;
1780 return buf;
1783 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1785 read_lock_bh(&tipc_net_lock);
1786 while (head) {
1787 struct bearer *b_ptr;
1788 struct node *n_ptr;
1789 struct link *l_ptr;
1790 struct sk_buff *crs;
1791 struct sk_buff *buf = head;
1792 struct tipc_msg *msg = buf_msg(buf);
1793 u32 seq_no = msg_seqno(msg);
1794 u32 ackd = msg_ack(msg);
1795 u32 released = 0;
1796 int type;
1798 b_ptr = (struct bearer *)tb_ptr;
1799 TIPC_SKB_CB(buf)->handle = b_ptr;
1801 head = head->next;
1802 if (unlikely(msg_version(msg) != TIPC_VERSION))
1803 goto cont;
1804 #if 0
1805 if (msg_user(msg) != LINK_PROTOCOL)
1806 #endif
1807 msg_dbg(msg,"<REC<");
1809 if (unlikely(msg_non_seq(msg))) {
1810 link_recv_non_seq(buf);
1811 continue;
1814 if (unlikely(!msg_short(msg) &&
1815 (msg_destnode(msg) != tipc_own_addr)))
1816 goto cont;
1818 n_ptr = tipc_node_find(msg_prevnode(msg));
1819 if (unlikely(!n_ptr))
1820 goto cont;
1822 tipc_node_lock(n_ptr);
1823 l_ptr = n_ptr->links[b_ptr->identity];
1824 if (unlikely(!l_ptr)) {
1825 tipc_node_unlock(n_ptr);
1826 goto cont;
1829 * Release acked messages
1831 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1832 if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1833 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1836 crs = l_ptr->first_out;
1837 while ((crs != l_ptr->next_out) &&
1838 less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1839 struct sk_buff *next = crs->next;
1841 buf_discard(crs);
1842 crs = next;
1843 released++;
1845 if (released) {
1846 l_ptr->first_out = crs;
1847 l_ptr->out_queue_size -= released;
1849 if (unlikely(l_ptr->next_out))
1850 tipc_link_push_queue(l_ptr);
1851 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1852 tipc_link_wakeup_ports(l_ptr, 0);
1853 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1854 l_ptr->stats.sent_acks++;
1855 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1858 protocol_check:
1859 if (likely(link_working_working(l_ptr))) {
1860 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1861 l_ptr->next_in_no++;
1862 if (unlikely(l_ptr->oldest_deferred_in))
1863 head = link_insert_deferred_queue(l_ptr,
1864 head);
1865 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1866 deliver:
1867 if (likely(msg_isdata(msg))) {
1868 tipc_node_unlock(n_ptr);
1869 tipc_port_recv_msg(buf);
1870 continue;
1872 switch (msg_user(msg)) {
1873 case MSG_BUNDLER:
1874 l_ptr->stats.recv_bundles++;
1875 l_ptr->stats.recv_bundled +=
1876 msg_msgcnt(msg);
1877 tipc_node_unlock(n_ptr);
1878 tipc_link_recv_bundle(buf);
1879 continue;
1880 case ROUTE_DISTRIBUTOR:
1881 tipc_node_unlock(n_ptr);
1882 tipc_cltr_recv_routing_table(buf);
1883 continue;
1884 case NAME_DISTRIBUTOR:
1885 tipc_node_unlock(n_ptr);
1886 tipc_named_recv(buf);
1887 continue;
1888 case CONN_MANAGER:
1889 tipc_node_unlock(n_ptr);
1890 tipc_port_recv_proto_msg(buf);
1891 continue;
1892 case MSG_FRAGMENTER:
1893 l_ptr->stats.recv_fragments++;
1894 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1895 &buf, &msg)) {
1896 l_ptr->stats.recv_fragmented++;
1897 goto deliver;
1899 break;
1900 case CHANGEOVER_PROTOCOL:
1901 type = msg_type(msg);
1902 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1903 msg = buf_msg(buf);
1904 seq_no = msg_seqno(msg);
1905 TIPC_SKB_CB(buf)->handle
1906 = b_ptr;
1907 if (type == ORIGINAL_MSG)
1908 goto deliver;
1909 goto protocol_check;
1911 break;
1914 tipc_node_unlock(n_ptr);
1915 tipc_net_route_msg(buf);
1916 continue;
1918 link_handle_out_of_seq_msg(l_ptr, buf);
1919 head = link_insert_deferred_queue(l_ptr, head);
1920 tipc_node_unlock(n_ptr);
1921 continue;
1924 if (msg_user(msg) == LINK_PROTOCOL) {
1925 link_recv_proto_msg(l_ptr, buf);
1926 head = link_insert_deferred_queue(l_ptr, head);
1927 tipc_node_unlock(n_ptr);
1928 continue;
1930 msg_dbg(msg,"NSEQ<REC<");
1931 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1933 if (link_working_working(l_ptr)) {
1934 /* Re-insert in front of queue */
1935 msg_dbg(msg,"RECV-REINS:");
1936 buf->next = head;
1937 head = buf;
1938 tipc_node_unlock(n_ptr);
1939 continue;
1941 tipc_node_unlock(n_ptr);
1942 cont:
1943 buf_discard(buf);
1945 read_unlock_bh(&tipc_net_lock);
1949 * link_defer_buf(): Sort a received out-of-sequence packet
1950 * into the deferred reception queue.
1951 * Returns the increase of the queue length,i.e. 0 or 1
1954 u32 tipc_link_defer_pkt(struct sk_buff **head,
1955 struct sk_buff **tail,
1956 struct sk_buff *buf)
1958 struct sk_buff *prev = NULL;
1959 struct sk_buff *crs = *head;
1960 u32 seq_no = msg_seqno(buf_msg(buf));
1962 buf->next = NULL;
1964 /* Empty queue ? */
1965 if (*head == NULL) {
1966 *head = *tail = buf;
1967 return 1;
1970 /* Last ? */
1971 if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1972 (*tail)->next = buf;
1973 *tail = buf;
1974 return 1;
1977 /* Scan through queue and sort it in */
1978 do {
1979 struct tipc_msg *msg = buf_msg(crs);
1981 if (less(seq_no, msg_seqno(msg))) {
1982 buf->next = crs;
1983 if (prev)
1984 prev->next = buf;
1985 else
1986 *head = buf;
1987 return 1;
1989 if (seq_no == msg_seqno(msg)) {
1990 break;
1992 prev = crs;
1993 crs = crs->next;
1995 while (crs);
1997 /* Message is a duplicate of an existing message */
1999 buf_discard(buf);
2000 return 0;
2003 /**
2004 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2007 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2008 struct sk_buff *buf)
2010 u32 seq_no = msg_seqno(buf_msg(buf));
2012 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2013 link_recv_proto_msg(l_ptr, buf);
2014 return;
2017 dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2018 seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2020 /* Record OOS packet arrival (force mismatch on next timeout) */
2022 l_ptr->checkpoint--;
2025 * Discard packet if a duplicate; otherwise add it to deferred queue
2026 * and notify peer of gap as per protocol specification
2029 if (less(seq_no, mod(l_ptr->next_in_no))) {
2030 l_ptr->stats.duplicates++;
2031 buf_discard(buf);
2032 return;
2035 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2036 &l_ptr->newest_deferred_in, buf)) {
2037 l_ptr->deferred_inqueue_sz++;
2038 l_ptr->stats.deferred_recv++;
2039 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2040 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2041 } else
2042 l_ptr->stats.duplicates++;
2046 * Send protocol message to the other endpoint.
2048 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2049 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2051 struct sk_buff *buf = NULL;
2052 struct tipc_msg *msg = l_ptr->pmsg;
2053 u32 msg_size = sizeof(l_ptr->proto_msg);
2055 if (link_blocked(l_ptr))
2056 return;
2057 msg_set_type(msg, msg_typ);
2058 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2059 msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2060 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2062 if (msg_typ == STATE_MSG) {
2063 u32 next_sent = mod(l_ptr->next_out_no);
2065 if (!tipc_link_is_up(l_ptr))
2066 return;
2067 if (l_ptr->next_out)
2068 next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2069 msg_set_next_sent(msg, next_sent);
2070 if (l_ptr->oldest_deferred_in) {
2071 u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2072 gap = mod(rec - mod(l_ptr->next_in_no));
2074 msg_set_seq_gap(msg, gap);
2075 if (gap)
2076 l_ptr->stats.sent_nacks++;
2077 msg_set_link_tolerance(msg, tolerance);
2078 msg_set_linkprio(msg, priority);
2079 msg_set_max_pkt(msg, ack_mtu);
2080 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2081 msg_set_probe(msg, probe_msg != 0);
2082 if (probe_msg) {
2083 u32 mtu = l_ptr->max_pkt;
2085 if ((mtu < l_ptr->max_pkt_target) &&
2086 link_working_working(l_ptr) &&
2087 l_ptr->fsm_msg_cnt) {
2088 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2089 if (l_ptr->max_pkt_probes == 10) {
2090 l_ptr->max_pkt_target = (msg_size - 4);
2091 l_ptr->max_pkt_probes = 0;
2092 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2094 l_ptr->max_pkt_probes++;
2097 l_ptr->stats.sent_probes++;
2099 l_ptr->stats.sent_states++;
2100 } else { /* RESET_MSG or ACTIVATE_MSG */
2101 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2102 msg_set_seq_gap(msg, 0);
2103 msg_set_next_sent(msg, 1);
2104 msg_set_link_tolerance(msg, l_ptr->tolerance);
2105 msg_set_linkprio(msg, l_ptr->priority);
2106 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2109 if (tipc_node_has_redundant_links(l_ptr->owner)) {
2110 msg_set_redundant_link(msg);
2111 } else {
2112 msg_clear_redundant_link(msg);
2114 msg_set_linkprio(msg, l_ptr->priority);
2116 /* Ensure sequence number will not fit : */
2118 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2120 /* Congestion? */
2122 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2123 if (!l_ptr->proto_msg_queue) {
2124 l_ptr->proto_msg_queue =
2125 buf_acquire(sizeof(l_ptr->proto_msg));
2127 buf = l_ptr->proto_msg_queue;
2128 if (!buf)
2129 return;
2130 memcpy(buf->data, (unchar *)msg, sizeof(l_ptr->proto_msg));
2131 return;
2133 msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2135 /* Message can be sent */
2137 msg_dbg(msg, ">>");
2139 buf = buf_acquire(msg_size);
2140 if (!buf)
2141 return;
2143 memcpy(buf->data, (unchar *)msg, sizeof(l_ptr->proto_msg));
2144 msg_set_size(buf_msg(buf), msg_size);
2146 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2147 l_ptr->unacked_window = 0;
2148 buf_discard(buf);
2149 return;
2152 /* New congestion */
2153 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2154 l_ptr->proto_msg_queue = buf;
2155 l_ptr->stats.bearer_congs++;
2159 * Receive protocol message :
2160 * Note that network plane id propagates through the network, and may
2161 * change at any time. The node with lowest address rules
2164 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2166 u32 rec_gap = 0;
2167 u32 max_pkt_info;
2168 u32 max_pkt_ack;
2169 u32 msg_tol;
2170 struct tipc_msg *msg = buf_msg(buf);
2172 dbg("AT(%u):", jiffies_to_msecs(jiffies));
2173 msg_dbg(msg, "<<");
2174 if (link_blocked(l_ptr))
2175 goto exit;
2177 /* record unnumbered packet arrival (force mismatch on next timeout) */
2179 l_ptr->checkpoint--;
2181 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2182 if (tipc_own_addr > msg_prevnode(msg))
2183 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2185 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2187 switch (msg_type(msg)) {
2189 case RESET_MSG:
2190 if (!link_working_unknown(l_ptr) && l_ptr->peer_session) {
2191 if (msg_session(msg) == l_ptr->peer_session) {
2192 dbg("Duplicate RESET: %u<->%u\n",
2193 msg_session(msg), l_ptr->peer_session);
2194 break; /* duplicate: ignore */
2197 /* fall thru' */
2198 case ACTIVATE_MSG:
2199 /* Update link settings according other endpoint's values */
2201 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2203 if ((msg_tol = msg_link_tolerance(msg)) &&
2204 (msg_tol > l_ptr->tolerance))
2205 link_set_supervision_props(l_ptr, msg_tol);
2207 if (msg_linkprio(msg) > l_ptr->priority)
2208 l_ptr->priority = msg_linkprio(msg);
2210 max_pkt_info = msg_max_pkt(msg);
2211 if (max_pkt_info) {
2212 if (max_pkt_info < l_ptr->max_pkt_target)
2213 l_ptr->max_pkt_target = max_pkt_info;
2214 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2215 l_ptr->max_pkt = l_ptr->max_pkt_target;
2216 } else {
2217 l_ptr->max_pkt = l_ptr->max_pkt_target;
2219 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2221 link_state_event(l_ptr, msg_type(msg));
2223 l_ptr->peer_session = msg_session(msg);
2224 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2226 /* Synchronize broadcast sequence numbers */
2227 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2228 l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2230 break;
2231 case STATE_MSG:
2233 if ((msg_tol = msg_link_tolerance(msg)))
2234 link_set_supervision_props(l_ptr, msg_tol);
2236 if (msg_linkprio(msg) &&
2237 (msg_linkprio(msg) != l_ptr->priority)) {
2238 warn("Resetting link <%s>, priority change %u->%u\n",
2239 l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2240 l_ptr->priority = msg_linkprio(msg);
2241 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2242 break;
2244 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2245 l_ptr->stats.recv_states++;
2246 if (link_reset_unknown(l_ptr))
2247 break;
2249 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2250 rec_gap = mod(msg_next_sent(msg) -
2251 mod(l_ptr->next_in_no));
2254 max_pkt_ack = msg_max_pkt(msg);
2255 if (max_pkt_ack > l_ptr->max_pkt) {
2256 dbg("Link <%s> updated MTU %u -> %u\n",
2257 l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2258 l_ptr->max_pkt = max_pkt_ack;
2259 l_ptr->max_pkt_probes = 0;
2262 max_pkt_ack = 0;
2263 if (msg_probe(msg)) {
2264 l_ptr->stats.recv_probes++;
2265 if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2266 max_pkt_ack = msg_size(msg);
2270 /* Protocol message before retransmits, reduce loss risk */
2272 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2274 if (rec_gap || (msg_probe(msg))) {
2275 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2276 0, rec_gap, 0, 0, max_pkt_ack);
2278 if (msg_seq_gap(msg)) {
2279 msg_dbg(msg, "With Gap:");
2280 l_ptr->stats.recv_nacks++;
2281 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2282 msg_seq_gap(msg));
2284 break;
2285 default:
2286 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2288 exit:
2289 buf_discard(buf);
2294 * tipc_link_tunnel(): Send one message via a link belonging to
2295 * another bearer. Owner node is locked.
2297 void tipc_link_tunnel(struct link *l_ptr,
2298 struct tipc_msg *tunnel_hdr,
2299 struct tipc_msg *msg,
2300 u32 selector)
2302 struct link *tunnel;
2303 struct sk_buff *buf;
2304 u32 length = msg_size(msg);
2306 tunnel = l_ptr->owner->active_links[selector & 1];
2307 if (!tipc_link_is_up(tunnel)) {
2308 warn("Link changeover error, "
2309 "tunnel link no longer available\n");
2310 return;
2312 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2313 buf = buf_acquire(length + INT_H_SIZE);
2314 if (!buf) {
2315 warn("Link changeover error, "
2316 "unable to send tunnel msg\n");
2317 return;
2319 memcpy(buf->data, (unchar *)tunnel_hdr, INT_H_SIZE);
2320 memcpy(buf->data + INT_H_SIZE, (unchar *)msg, length);
2321 dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2322 msg_dbg(buf_msg(buf), ">SEND>");
2323 tipc_link_send_buf(tunnel, buf);
2329 * changeover(): Send whole message queue via the remaining link
2330 * Owner node is locked.
2333 void tipc_link_changeover(struct link *l_ptr)
2335 u32 msgcount = l_ptr->out_queue_size;
2336 struct sk_buff *crs = l_ptr->first_out;
2337 struct link *tunnel = l_ptr->owner->active_links[0];
2338 struct tipc_msg tunnel_hdr;
2339 int split_bundles;
2341 if (!tunnel)
2342 return;
2344 if (!l_ptr->owner->permit_changeover) {
2345 warn("Link changeover error, "
2346 "peer did not permit changeover\n");
2347 return;
2350 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2351 ORIGINAL_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2352 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2353 msg_set_msgcnt(&tunnel_hdr, msgcount);
2354 dbg("Link changeover requires %u tunnel messages\n", msgcount);
2356 if (!l_ptr->first_out) {
2357 struct sk_buff *buf;
2359 buf = buf_acquire(INT_H_SIZE);
2360 if (buf) {
2361 memcpy(buf->data, (unchar *)&tunnel_hdr, INT_H_SIZE);
2362 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2363 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2364 tunnel->b_ptr->net_plane);
2365 msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2366 tipc_link_send_buf(tunnel, buf);
2367 } else {
2368 warn("Link changeover error, "
2369 "unable to send changeover msg\n");
2371 return;
2374 split_bundles = (l_ptr->owner->active_links[0] !=
2375 l_ptr->owner->active_links[1]);
2377 while (crs) {
2378 struct tipc_msg *msg = buf_msg(crs);
2380 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2381 u32 msgcount = msg_msgcnt(msg);
2382 struct tipc_msg *m = msg_get_wrapped(msg);
2383 unchar* pos = (unchar*)m;
2385 while (msgcount--) {
2386 msg_set_seqno(m,msg_seqno(msg));
2387 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2388 msg_link_selector(m));
2389 pos += align(msg_size(m));
2390 m = (struct tipc_msg *)pos;
2392 } else {
2393 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2394 msg_link_selector(msg));
2396 crs = crs->next;
2400 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2402 struct sk_buff *iter;
2403 struct tipc_msg tunnel_hdr;
2405 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2406 DUPLICATE_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2407 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2408 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2409 iter = l_ptr->first_out;
2410 while (iter) {
2411 struct sk_buff *outbuf;
2412 struct tipc_msg *msg = buf_msg(iter);
2413 u32 length = msg_size(msg);
2415 if (msg_user(msg) == MSG_BUNDLER)
2416 msg_set_type(msg, CLOSED_MSG);
2417 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2418 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2419 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2420 outbuf = buf_acquire(length + INT_H_SIZE);
2421 if (outbuf == NULL) {
2422 warn("Link changeover error, "
2423 "unable to send duplicate msg\n");
2424 return;
2426 memcpy(outbuf->data, (unchar *)&tunnel_hdr, INT_H_SIZE);
2427 memcpy(outbuf->data + INT_H_SIZE, iter->data, length);
2428 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2429 tunnel->b_ptr->net_plane);
2430 msg_dbg(buf_msg(outbuf), ">SEND>");
2431 tipc_link_send_buf(tunnel, outbuf);
2432 if (!tipc_link_is_up(l_ptr))
2433 return;
2434 iter = iter->next;
2441 * buf_extract - extracts embedded TIPC message from another message
2442 * @skb: encapsulating message buffer
2443 * @from_pos: offset to extract from
2445 * Returns a new message buffer containing an embedded message. The
2446 * encapsulating message itself is left unchanged.
2449 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2451 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2452 u32 size = msg_size(msg);
2453 struct sk_buff *eb;
2455 eb = buf_acquire(size);
2456 if (eb)
2457 memcpy(eb->data, (unchar *)msg, size);
2458 return eb;
2462 * link_recv_changeover_msg(): Receive tunneled packet sent
2463 * via other link. Node is locked. Return extracted buffer.
2466 static int link_recv_changeover_msg(struct link **l_ptr,
2467 struct sk_buff **buf)
2469 struct sk_buff *tunnel_buf = *buf;
2470 struct link *dest_link;
2471 struct tipc_msg *msg;
2472 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2473 u32 msg_typ = msg_type(tunnel_msg);
2474 u32 msg_count = msg_msgcnt(tunnel_msg);
2476 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2477 if (!dest_link) {
2478 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2479 goto exit;
2481 if (dest_link == *l_ptr) {
2482 err("Unexpected changeover message on link <%s>\n",
2483 (*l_ptr)->name);
2484 goto exit;
2486 dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2487 (*l_ptr)->b_ptr->net_plane);
2488 *l_ptr = dest_link;
2489 msg = msg_get_wrapped(tunnel_msg);
2491 if (msg_typ == DUPLICATE_MSG) {
2492 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2493 msg_dbg(tunnel_msg, "DROP/<REC<");
2494 goto exit;
2496 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2497 if (*buf == NULL) {
2498 warn("Link changeover error, duplicate msg dropped\n");
2499 goto exit;
2501 msg_dbg(tunnel_msg, "TNL<REC<");
2502 buf_discard(tunnel_buf);
2503 return 1;
2506 /* First original message ?: */
2508 if (tipc_link_is_up(dest_link)) {
2509 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2510 info("Resetting link <%s>, changeover initiated by peer\n",
2511 dest_link->name);
2512 tipc_link_reset(dest_link);
2513 dest_link->exp_msg_count = msg_count;
2514 dbg("Expecting %u tunnelled messages\n", msg_count);
2515 if (!msg_count)
2516 goto exit;
2517 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2518 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2519 dest_link->exp_msg_count = msg_count;
2520 dbg("Expecting %u tunnelled messages\n", msg_count);
2521 if (!msg_count)
2522 goto exit;
2525 /* Receive original message */
2527 if (dest_link->exp_msg_count == 0) {
2528 warn("Link switchover error, "
2529 "got too many tunnelled messages\n");
2530 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2531 dbg_print_link(dest_link, "LINK:");
2532 goto exit;
2534 dest_link->exp_msg_count--;
2535 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2536 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2537 goto exit;
2538 } else {
2539 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2540 if (*buf != NULL) {
2541 msg_dbg(tunnel_msg, "TNL<REC<");
2542 buf_discard(tunnel_buf);
2543 return 1;
2544 } else {
2545 warn("Link changeover error, original msg dropped\n");
2548 exit:
2549 *buf = NULL;
2550 buf_discard(tunnel_buf);
2551 return 0;
2555 * Bundler functionality:
2557 void tipc_link_recv_bundle(struct sk_buff *buf)
2559 u32 msgcount = msg_msgcnt(buf_msg(buf));
2560 u32 pos = INT_H_SIZE;
2561 struct sk_buff *obuf;
2563 msg_dbg(buf_msg(buf), "<BNDL<: ");
2564 while (msgcount--) {
2565 obuf = buf_extract(buf, pos);
2566 if (obuf == NULL) {
2567 warn("Link unable to unbundle message(s)\n");
2568 break;
2570 pos += align(msg_size(buf_msg(obuf)));
2571 msg_dbg(buf_msg(obuf), " /");
2572 tipc_net_route_msg(obuf);
2574 buf_discard(buf);
2578 * Fragmentation/defragmentation:
2583 * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2584 * The buffer is complete, inclusive total message length.
2585 * Returns user data length.
2587 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2589 struct tipc_msg *inmsg = buf_msg(buf);
2590 struct tipc_msg fragm_hdr;
2591 u32 insize = msg_size(inmsg);
2592 u32 dsz = msg_data_sz(inmsg);
2593 unchar *crs = buf->data;
2594 u32 rest = insize;
2595 u32 pack_sz = link_max_pkt(l_ptr);
2596 u32 fragm_sz = pack_sz - INT_H_SIZE;
2597 u32 fragm_no = 1;
2598 u32 destaddr = msg_destnode(inmsg);
2600 if (msg_short(inmsg))
2601 destaddr = l_ptr->addr;
2603 if (msg_routed(inmsg))
2604 msg_set_prevnode(inmsg, tipc_own_addr);
2606 /* Prepare reusable fragment header: */
2608 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2609 TIPC_OK, INT_H_SIZE, destaddr);
2610 msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2611 msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2612 msg_set_fragm_no(&fragm_hdr, fragm_no);
2613 l_ptr->stats.sent_fragmented++;
2615 /* Chop up message: */
2617 while (rest > 0) {
2618 struct sk_buff *fragm;
2620 if (rest <= fragm_sz) {
2621 fragm_sz = rest;
2622 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2624 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2625 if (fragm == NULL) {
2626 warn("Link unable to fragment message\n");
2627 dsz = -ENOMEM;
2628 goto exit;
2630 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2631 memcpy(fragm->data, (unchar *)&fragm_hdr, INT_H_SIZE);
2632 memcpy(fragm->data + INT_H_SIZE, crs, fragm_sz);
2634 /* Send queued messages first, if any: */
2636 l_ptr->stats.sent_fragments++;
2637 tipc_link_send_buf(l_ptr, fragm);
2638 if (!tipc_link_is_up(l_ptr))
2639 return dsz;
2640 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2641 rest -= fragm_sz;
2642 crs += fragm_sz;
2643 msg_set_type(&fragm_hdr, FRAGMENT);
2645 exit:
2646 buf_discard(buf);
2647 return dsz;
2651 * A pending message being re-assembled must store certain values
2652 * to handle subsequent fragments correctly. The following functions
2653 * help storing these values in unused, available fields in the
2654 * pending message. This makes dynamic memory allocation unecessary.
2657 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2659 msg_set_seqno(buf_msg(buf), seqno);
2662 static u32 get_fragm_size(struct sk_buff *buf)
2664 return msg_ack(buf_msg(buf));
2667 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2669 msg_set_ack(buf_msg(buf), sz);
2672 static u32 get_expected_frags(struct sk_buff *buf)
2674 return msg_bcast_ack(buf_msg(buf));
2677 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2679 msg_set_bcast_ack(buf_msg(buf), exp);
2682 static u32 get_timer_cnt(struct sk_buff *buf)
2684 return msg_reroute_cnt(buf_msg(buf));
2687 static void incr_timer_cnt(struct sk_buff *buf)
2689 msg_incr_reroute_cnt(buf_msg(buf));
2693 * tipc_link_recv_fragment(): Called with node lock on. Returns
2694 * the reassembled buffer if message is complete.
2696 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2697 struct tipc_msg **m)
2699 struct sk_buff *prev = NULL;
2700 struct sk_buff *fbuf = *fb;
2701 struct tipc_msg *fragm = buf_msg(fbuf);
2702 struct sk_buff *pbuf = *pending;
2703 u32 long_msg_seq_no = msg_long_msgno(fragm);
2705 *fb = NULL;
2706 msg_dbg(fragm,"FRG<REC<");
2708 /* Is there an incomplete message waiting for this fragment? */
2710 while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no)
2711 || (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2712 prev = pbuf;
2713 pbuf = pbuf->next;
2716 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2717 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2718 u32 msg_sz = msg_size(imsg);
2719 u32 fragm_sz = msg_data_sz(fragm);
2720 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2721 u32 max = TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2722 if (msg_type(imsg) == TIPC_MCAST_MSG)
2723 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2724 if (msg_size(imsg) > max) {
2725 msg_dbg(fragm,"<REC<Oversized: ");
2726 buf_discard(fbuf);
2727 return 0;
2729 pbuf = buf_acquire(msg_size(imsg));
2730 if (pbuf != NULL) {
2731 pbuf->next = *pending;
2732 *pending = pbuf;
2733 memcpy(pbuf->data, (unchar *)imsg, msg_data_sz(fragm));
2735 /* Prepare buffer for subsequent fragments. */
2737 set_long_msg_seqno(pbuf, long_msg_seq_no);
2738 set_fragm_size(pbuf,fragm_sz);
2739 set_expected_frags(pbuf,exp_fragm_cnt - 1);
2740 } else {
2741 warn("Link unable to reassemble fragmented message\n");
2743 buf_discard(fbuf);
2744 return 0;
2745 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2746 u32 dsz = msg_data_sz(fragm);
2747 u32 fsz = get_fragm_size(pbuf);
2748 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2749 u32 exp_frags = get_expected_frags(pbuf) - 1;
2750 memcpy(pbuf->data + crs, msg_data(fragm), dsz);
2751 buf_discard(fbuf);
2753 /* Is message complete? */
2755 if (exp_frags == 0) {
2756 if (prev)
2757 prev->next = pbuf->next;
2758 else
2759 *pending = pbuf->next;
2760 msg_reset_reroute_cnt(buf_msg(pbuf));
2761 *fb = pbuf;
2762 *m = buf_msg(pbuf);
2763 return 1;
2765 set_expected_frags(pbuf,exp_frags);
2766 return 0;
2768 dbg(" Discarding orphan fragment %x\n",fbuf);
2769 msg_dbg(fragm,"ORPHAN:");
2770 dbg("Pending long buffers:\n");
2771 dbg_print_buf_chain(*pending);
2772 buf_discard(fbuf);
2773 return 0;
2777 * link_check_defragm_bufs - flush stale incoming message fragments
2778 * @l_ptr: pointer to link
2781 static void link_check_defragm_bufs(struct link *l_ptr)
2783 struct sk_buff *prev = NULL;
2784 struct sk_buff *next = NULL;
2785 struct sk_buff *buf = l_ptr->defragm_buf;
2787 if (!buf)
2788 return;
2789 if (!link_working_working(l_ptr))
2790 return;
2791 while (buf) {
2792 u32 cnt = get_timer_cnt(buf);
2794 next = buf->next;
2795 if (cnt < 4) {
2796 incr_timer_cnt(buf);
2797 prev = buf;
2798 } else {
2799 dbg(" Discarding incomplete long buffer\n");
2800 msg_dbg(buf_msg(buf), "LONG:");
2801 dbg_print_link(l_ptr, "curr:");
2802 dbg("Pending long buffers:\n");
2803 dbg_print_buf_chain(l_ptr->defragm_buf);
2804 if (prev)
2805 prev->next = buf->next;
2806 else
2807 l_ptr->defragm_buf = buf->next;
2808 buf_discard(buf);
2810 buf = next;
2816 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2818 l_ptr->tolerance = tolerance;
2819 l_ptr->continuity_interval =
2820 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2821 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2825 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2827 /* Data messages from this node, inclusive FIRST_FRAGM */
2828 l_ptr->queue_limit[DATA_LOW] = window;
2829 l_ptr->queue_limit[DATA_MEDIUM] = (window / 3) * 4;
2830 l_ptr->queue_limit[DATA_HIGH] = (window / 3) * 5;
2831 l_ptr->queue_limit[DATA_CRITICAL] = (window / 3) * 6;
2832 /* Transiting data messages,inclusive FIRST_FRAGM */
2833 l_ptr->queue_limit[DATA_LOW + 4] = 300;
2834 l_ptr->queue_limit[DATA_MEDIUM + 4] = 600;
2835 l_ptr->queue_limit[DATA_HIGH + 4] = 900;
2836 l_ptr->queue_limit[DATA_CRITICAL + 4] = 1200;
2837 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2838 l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2839 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2840 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2841 /* FRAGMENT and LAST_FRAGMENT packets */
2842 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2846 * link_find_link - locate link by name
2847 * @name - ptr to link name string
2848 * @node - ptr to area to be filled with ptr to associated node
2850 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2851 * this also prevents link deletion.
2853 * Returns pointer to link (or 0 if invalid link name).
2856 static struct link *link_find_link(const char *name, struct node **node)
2858 struct link_name link_name_parts;
2859 struct bearer *b_ptr;
2860 struct link *l_ptr;
2862 if (!link_name_validate(name, &link_name_parts))
2863 return NULL;
2865 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2866 if (!b_ptr)
2867 return NULL;
2869 *node = tipc_node_find(link_name_parts.addr_peer);
2870 if (!*node)
2871 return NULL;
2873 l_ptr = (*node)->links[b_ptr->identity];
2874 if (!l_ptr || strcmp(l_ptr->name, name))
2875 return NULL;
2877 return l_ptr;
2880 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2881 u16 cmd)
2883 struct tipc_link_config *args;
2884 u32 new_value;
2885 struct link *l_ptr;
2886 struct node *node;
2887 int res;
2889 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2890 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2892 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2893 new_value = ntohl(args->value);
2895 if (!strcmp(args->name, tipc_bclink_name)) {
2896 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2897 (tipc_bclink_set_queue_limits(new_value) == 0))
2898 return tipc_cfg_reply_none();
2899 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2900 " (cannot change setting on broadcast link)");
2903 read_lock_bh(&tipc_net_lock);
2904 l_ptr = link_find_link(args->name, &node);
2905 if (!l_ptr) {
2906 read_unlock_bh(&tipc_net_lock);
2907 return tipc_cfg_reply_error_string("link not found");
2910 tipc_node_lock(node);
2911 res = -EINVAL;
2912 switch (cmd) {
2913 case TIPC_CMD_SET_LINK_TOL:
2914 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2915 (new_value <= TIPC_MAX_LINK_TOL)) {
2916 link_set_supervision_props(l_ptr, new_value);
2917 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2918 0, 0, new_value, 0, 0);
2919 res = TIPC_OK;
2921 break;
2922 case TIPC_CMD_SET_LINK_PRI:
2923 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2924 (new_value <= TIPC_MAX_LINK_PRI)) {
2925 l_ptr->priority = new_value;
2926 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2927 0, 0, 0, new_value, 0);
2928 res = TIPC_OK;
2930 break;
2931 case TIPC_CMD_SET_LINK_WINDOW:
2932 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2933 (new_value <= TIPC_MAX_LINK_WIN)) {
2934 tipc_link_set_queue_limits(l_ptr, new_value);
2935 res = TIPC_OK;
2937 break;
2939 tipc_node_unlock(node);
2941 read_unlock_bh(&tipc_net_lock);
2942 if (res)
2943 return tipc_cfg_reply_error_string("cannot change link setting");
2945 return tipc_cfg_reply_none();
2949 * link_reset_statistics - reset link statistics
2950 * @l_ptr: pointer to link
2953 static void link_reset_statistics(struct link *l_ptr)
2955 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2956 l_ptr->stats.sent_info = l_ptr->next_out_no;
2957 l_ptr->stats.recv_info = l_ptr->next_in_no;
2960 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2962 char *link_name;
2963 struct link *l_ptr;
2964 struct node *node;
2966 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2967 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2969 link_name = (char *)TLV_DATA(req_tlv_area);
2970 if (!strcmp(link_name, tipc_bclink_name)) {
2971 if (tipc_bclink_reset_stats())
2972 return tipc_cfg_reply_error_string("link not found");
2973 return tipc_cfg_reply_none();
2976 read_lock_bh(&tipc_net_lock);
2977 l_ptr = link_find_link(link_name, &node);
2978 if (!l_ptr) {
2979 read_unlock_bh(&tipc_net_lock);
2980 return tipc_cfg_reply_error_string("link not found");
2983 tipc_node_lock(node);
2984 link_reset_statistics(l_ptr);
2985 tipc_node_unlock(node);
2986 read_unlock_bh(&tipc_net_lock);
2987 return tipc_cfg_reply_none();
2991 * percent - convert count to a percentage of total (rounding up or down)
2994 static u32 percent(u32 count, u32 total)
2996 return (count * 100 + (total / 2)) / total;
3000 * tipc_link_stats - print link statistics
3001 * @name: link name
3002 * @buf: print buffer area
3003 * @buf_size: size of print buffer area
3005 * Returns length of print buffer data string (or 0 if error)
3008 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3010 struct print_buf pb;
3011 struct link *l_ptr;
3012 struct node *node;
3013 char *status;
3014 u32 profile_total = 0;
3016 if (!strcmp(name, tipc_bclink_name))
3017 return tipc_bclink_stats(buf, buf_size);
3019 tipc_printbuf_init(&pb, buf, buf_size);
3021 read_lock_bh(&tipc_net_lock);
3022 l_ptr = link_find_link(name, &node);
3023 if (!l_ptr) {
3024 read_unlock_bh(&tipc_net_lock);
3025 return 0;
3027 tipc_node_lock(node);
3029 if (tipc_link_is_active(l_ptr))
3030 status = "ACTIVE";
3031 else if (tipc_link_is_up(l_ptr))
3032 status = "STANDBY";
3033 else
3034 status = "DEFUNCT";
3035 tipc_printf(&pb, "Link <%s>\n"
3036 " %s MTU:%u Priority:%u Tolerance:%u ms"
3037 " Window:%u packets\n",
3038 l_ptr->name, status, link_max_pkt(l_ptr),
3039 l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3040 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3041 l_ptr->next_in_no - l_ptr->stats.recv_info,
3042 l_ptr->stats.recv_fragments,
3043 l_ptr->stats.recv_fragmented,
3044 l_ptr->stats.recv_bundles,
3045 l_ptr->stats.recv_bundled);
3046 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3047 l_ptr->next_out_no - l_ptr->stats.sent_info,
3048 l_ptr->stats.sent_fragments,
3049 l_ptr->stats.sent_fragmented,
3050 l_ptr->stats.sent_bundles,
3051 l_ptr->stats.sent_bundled);
3052 profile_total = l_ptr->stats.msg_length_counts;
3053 if (!profile_total)
3054 profile_total = 1;
3055 tipc_printf(&pb, " TX profile sample:%u packets average:%u octets\n"
3056 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3057 "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3058 l_ptr->stats.msg_length_counts,
3059 l_ptr->stats.msg_lengths_total / profile_total,
3060 percent(l_ptr->stats.msg_length_profile[0], profile_total),
3061 percent(l_ptr->stats.msg_length_profile[1], profile_total),
3062 percent(l_ptr->stats.msg_length_profile[2], profile_total),
3063 percent(l_ptr->stats.msg_length_profile[3], profile_total),
3064 percent(l_ptr->stats.msg_length_profile[4], profile_total),
3065 percent(l_ptr->stats.msg_length_profile[5], profile_total),
3066 percent(l_ptr->stats.msg_length_profile[6], profile_total));
3067 tipc_printf(&pb, " RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3068 l_ptr->stats.recv_states,
3069 l_ptr->stats.recv_probes,
3070 l_ptr->stats.recv_nacks,
3071 l_ptr->stats.deferred_recv,
3072 l_ptr->stats.duplicates);
3073 tipc_printf(&pb, " TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3074 l_ptr->stats.sent_states,
3075 l_ptr->stats.sent_probes,
3076 l_ptr->stats.sent_nacks,
3077 l_ptr->stats.sent_acks,
3078 l_ptr->stats.retransmitted);
3079 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
3080 l_ptr->stats.bearer_congs,
3081 l_ptr->stats.link_congs,
3082 l_ptr->stats.max_queue_sz,
3083 l_ptr->stats.queue_sz_counts
3084 ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3085 : 0);
3087 tipc_node_unlock(node);
3088 read_unlock_bh(&tipc_net_lock);
3089 return tipc_printbuf_validate(&pb);
3092 #define MAX_LINK_STATS_INFO 2000
3094 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3096 struct sk_buff *buf;
3097 struct tlv_desc *rep_tlv;
3098 int str_len;
3100 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3101 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3103 buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3104 if (!buf)
3105 return NULL;
3107 rep_tlv = (struct tlv_desc *)buf->data;
3109 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3110 (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3111 if (!str_len) {
3112 buf_discard(buf);
3113 return tipc_cfg_reply_error_string("link not found");
3116 skb_put(buf, TLV_SPACE(str_len));
3117 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3119 return buf;
3122 #if 0
3123 int link_control(const char *name, u32 op, u32 val)
3125 int res = -EINVAL;
3126 struct link *l_ptr;
3127 u32 bearer_id;
3128 struct node * node;
3129 u32 a;
3131 a = link_name2addr(name, &bearer_id);
3132 read_lock_bh(&tipc_net_lock);
3133 node = tipc_node_find(a);
3134 if (node) {
3135 tipc_node_lock(node);
3136 l_ptr = node->links[bearer_id];
3137 if (l_ptr) {
3138 if (op == TIPC_REMOVE_LINK) {
3139 struct bearer *b_ptr = l_ptr->b_ptr;
3140 spin_lock_bh(&b_ptr->publ.lock);
3141 tipc_link_delete(l_ptr);
3142 spin_unlock_bh(&b_ptr->publ.lock);
3144 if (op == TIPC_CMD_BLOCK_LINK) {
3145 tipc_link_reset(l_ptr);
3146 l_ptr->blocked = 1;
3148 if (op == TIPC_CMD_UNBLOCK_LINK) {
3149 l_ptr->blocked = 0;
3151 res = TIPC_OK;
3153 tipc_node_unlock(node);
3155 read_unlock_bh(&tipc_net_lock);
3156 return res;
3158 #endif
3161 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3162 * @dest: network address of destination node
3163 * @selector: used to select from set of active links
3165 * If no active link can be found, uses default maximum packet size.
3168 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3170 struct node *n_ptr;
3171 struct link *l_ptr;
3172 u32 res = MAX_PKT_DEFAULT;
3174 if (dest == tipc_own_addr)
3175 return MAX_MSG_SIZE;
3177 read_lock_bh(&tipc_net_lock);
3178 n_ptr = tipc_node_select(dest, selector);
3179 if (n_ptr) {
3180 tipc_node_lock(n_ptr);
3181 l_ptr = n_ptr->active_links[selector & 1];
3182 if (l_ptr)
3183 res = link_max_pkt(l_ptr);
3184 tipc_node_unlock(n_ptr);
3186 read_unlock_bh(&tipc_net_lock);
3187 return res;
3190 #if 0
3191 static void link_dump_rec_queue(struct link *l_ptr)
3193 struct sk_buff *crs;
3195 if (!l_ptr->oldest_deferred_in) {
3196 info("Reception queue empty\n");
3197 return;
3199 info("Contents of Reception queue:\n");
3200 crs = l_ptr->oldest_deferred_in;
3201 while (crs) {
3202 if (crs->data == (void *)0x0000a3a3) {
3203 info("buffer %x invalid\n", crs);
3204 return;
3206 msg_dbg(buf_msg(crs), "In rec queue: \n");
3207 crs = crs->next;
3210 #endif
3212 static void link_dump_send_queue(struct link *l_ptr)
3214 if (l_ptr->next_out) {
3215 info("\nContents of unsent queue:\n");
3216 dbg_print_buf_chain(l_ptr->next_out);
3218 info("\nContents of send queue:\n");
3219 if (l_ptr->first_out) {
3220 dbg_print_buf_chain(l_ptr->first_out);
3222 info("Empty send queue\n");
3225 static void link_print(struct link *l_ptr, struct print_buf *buf,
3226 const char *str)
3228 tipc_printf(buf, str);
3229 if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3230 return;
3231 tipc_printf(buf, "Link %x<%s>:",
3232 l_ptr->addr, l_ptr->b_ptr->publ.name);
3233 tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3234 tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3235 tipc_printf(buf, "SQUE");
3236 if (l_ptr->first_out) {
3237 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3238 if (l_ptr->next_out)
3239 tipc_printf(buf, "%u..",
3240 msg_seqno(buf_msg(l_ptr->next_out)));
3241 tipc_printf(buf, "%u]",
3242 msg_seqno(buf_msg
3243 (l_ptr->last_out)), l_ptr->out_queue_size);
3244 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3245 msg_seqno(buf_msg(l_ptr->first_out)))
3246 != (l_ptr->out_queue_size - 1))
3247 || (l_ptr->last_out->next != 0)) {
3248 tipc_printf(buf, "\nSend queue inconsistency\n");
3249 tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3250 tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3251 tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3252 link_dump_send_queue(l_ptr);
3254 } else
3255 tipc_printf(buf, "[]");
3256 tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3257 if (l_ptr->oldest_deferred_in) {
3258 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3259 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3260 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3261 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3262 tipc_printf(buf, ":RQSIZ(%u)",
3263 l_ptr->deferred_inqueue_sz);
3266 if (link_working_unknown(l_ptr))
3267 tipc_printf(buf, ":WU");
3268 if (link_reset_reset(l_ptr))
3269 tipc_printf(buf, ":RR");
3270 if (link_reset_unknown(l_ptr))
3271 tipc_printf(buf, ":RU");
3272 if (link_working_working(l_ptr))
3273 tipc_printf(buf, ":WW");
3274 tipc_printf(buf, "\n");