libata bugfix: HDIO_DRIVE_TASK
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / tipc / link.c
blob1bb983c8130b32be1b70e776b04ee573f6530c42
1 /*
2 * net/tipc/link.c: TIPC link code
3 *
4 * Copyright (c) 1996-2006, Ericsson AB
5 * Copyright (c) 2004-2006, 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 (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 l_ptr->addr = peer;
427 if_name = strchr(b_ptr->publ.name, ':') + 1;
428 sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:",
429 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
430 tipc_node(tipc_own_addr),
431 if_name,
432 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
433 /* note: peer i/f is appended to link name by reset/activate */
434 memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
435 k_init_timer(&l_ptr->timer, (Handler)link_timeout, (unsigned long)l_ptr);
436 list_add_tail(&l_ptr->link_list, &b_ptr->links);
437 l_ptr->checkpoint = 1;
438 l_ptr->b_ptr = b_ptr;
439 link_set_supervision_props(l_ptr, b_ptr->media->tolerance);
440 l_ptr->state = RESET_UNKNOWN;
442 l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
443 msg = l_ptr->pmsg;
444 msg_init(msg, LINK_PROTOCOL, RESET_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
445 msg_set_size(msg, sizeof(l_ptr->proto_msg));
446 msg_set_session(msg, tipc_random);
447 msg_set_bearer_id(msg, b_ptr->identity);
448 strcpy((char *)msg_data(msg), if_name);
450 l_ptr->priority = b_ptr->priority;
451 tipc_link_set_queue_limits(l_ptr, b_ptr->media->window);
453 link_init_max_pkt(l_ptr);
455 l_ptr->next_out_no = 1;
456 INIT_LIST_HEAD(&l_ptr->waiting_ports);
458 link_reset_statistics(l_ptr);
460 l_ptr->owner = tipc_node_attach_link(l_ptr);
461 if (!l_ptr->owner) {
462 kfree(l_ptr);
463 return NULL;
466 if (LINK_LOG_BUF_SIZE) {
467 char *pb = kmalloc(LINK_LOG_BUF_SIZE, GFP_ATOMIC);
469 if (!pb) {
470 kfree(l_ptr);
471 warn("Link creation failed, no memory for print buffer\n");
472 return NULL;
474 tipc_printbuf_init(&l_ptr->print_buf, pb, LINK_LOG_BUF_SIZE);
477 tipc_k_signal((Handler)tipc_link_start, (unsigned long)l_ptr);
479 dbg("tipc_link_create(): tolerance = %u,cont intv = %u, abort_limit = %u\n",
480 l_ptr->tolerance, l_ptr->continuity_interval, l_ptr->abort_limit);
482 return l_ptr;
485 /**
486 * tipc_link_delete - delete a link
487 * @l_ptr: pointer to link
489 * Note: 'tipc_net_lock' is write_locked, bearer is locked.
490 * This routine must not grab the node lock until after link timer cancellation
491 * to avoid a potential deadlock situation.
494 void tipc_link_delete(struct link *l_ptr)
496 if (!l_ptr) {
497 err("Attempt to delete non-existent link\n");
498 return;
501 dbg("tipc_link_delete()\n");
503 k_cancel_timer(&l_ptr->timer);
505 tipc_node_lock(l_ptr->owner);
506 tipc_link_reset(l_ptr);
507 tipc_node_detach_link(l_ptr->owner, l_ptr);
508 tipc_link_stop(l_ptr);
509 list_del_init(&l_ptr->link_list);
510 if (LINK_LOG_BUF_SIZE)
511 kfree(l_ptr->print_buf.buf);
512 tipc_node_unlock(l_ptr->owner);
513 k_term_timer(&l_ptr->timer);
514 kfree(l_ptr);
517 void tipc_link_start(struct link *l_ptr)
519 dbg("tipc_link_start %x\n", l_ptr);
520 link_state_event(l_ptr, STARTING_EVT);
524 * link_schedule_port - schedule port for deferred sending
525 * @l_ptr: pointer to link
526 * @origport: reference to sending port
527 * @sz: amount of data to be sent
529 * Schedules port for renewed sending of messages after link congestion
530 * has abated.
533 static int link_schedule_port(struct link *l_ptr, u32 origport, u32 sz)
535 struct port *p_ptr;
537 spin_lock_bh(&tipc_port_list_lock);
538 p_ptr = tipc_port_lock(origport);
539 if (p_ptr) {
540 if (!p_ptr->wakeup)
541 goto exit;
542 if (!list_empty(&p_ptr->wait_list))
543 goto exit;
544 p_ptr->congested_link = l_ptr;
545 p_ptr->publ.congested = 1;
546 p_ptr->waiting_pkts = 1 + ((sz - 1) / link_max_pkt(l_ptr));
547 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
548 l_ptr->stats.link_congs++;
549 exit:
550 tipc_port_unlock(p_ptr);
552 spin_unlock_bh(&tipc_port_list_lock);
553 return -ELINKCONG;
556 void tipc_link_wakeup_ports(struct link *l_ptr, int all)
558 struct port *p_ptr;
559 struct port *temp_p_ptr;
560 int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
562 if (all)
563 win = 100000;
564 if (win <= 0)
565 return;
566 if (!spin_trylock_bh(&tipc_port_list_lock))
567 return;
568 if (link_congested(l_ptr))
569 goto exit;
570 list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
571 wait_list) {
572 if (win <= 0)
573 break;
574 list_del_init(&p_ptr->wait_list);
575 p_ptr->congested_link = NULL;
576 spin_lock_bh(p_ptr->publ.lock);
577 p_ptr->publ.congested = 0;
578 p_ptr->wakeup(&p_ptr->publ);
579 win -= p_ptr->waiting_pkts;
580 spin_unlock_bh(p_ptr->publ.lock);
583 exit:
584 spin_unlock_bh(&tipc_port_list_lock);
587 /**
588 * link_release_outqueue - purge link's outbound message queue
589 * @l_ptr: pointer to link
592 static void link_release_outqueue(struct link *l_ptr)
594 struct sk_buff *buf = l_ptr->first_out;
595 struct sk_buff *next;
597 while (buf) {
598 next = buf->next;
599 buf_discard(buf);
600 buf = next;
602 l_ptr->first_out = NULL;
603 l_ptr->out_queue_size = 0;
607 * tipc_link_reset_fragments - purge link's inbound message fragments queue
608 * @l_ptr: pointer to link
611 void tipc_link_reset_fragments(struct link *l_ptr)
613 struct sk_buff *buf = l_ptr->defragm_buf;
614 struct sk_buff *next;
616 while (buf) {
617 next = buf->next;
618 buf_discard(buf);
619 buf = next;
621 l_ptr->defragm_buf = NULL;
624 /**
625 * tipc_link_stop - purge all inbound and outbound messages associated with link
626 * @l_ptr: pointer to link
629 void tipc_link_stop(struct link *l_ptr)
631 struct sk_buff *buf;
632 struct sk_buff *next;
634 buf = l_ptr->oldest_deferred_in;
635 while (buf) {
636 next = buf->next;
637 buf_discard(buf);
638 buf = next;
641 buf = l_ptr->first_out;
642 while (buf) {
643 next = buf->next;
644 buf_discard(buf);
645 buf = next;
648 tipc_link_reset_fragments(l_ptr);
650 buf_discard(l_ptr->proto_msg_queue);
651 l_ptr->proto_msg_queue = NULL;
654 #if 0
656 /* LINK EVENT CODE IS NOT SUPPORTED AT PRESENT */
658 static void link_recv_event(struct link_event *ev)
660 ev->fcn(ev->addr, ev->name, ev->up);
661 kfree(ev);
664 static void link_send_event(void (*fcn)(u32 a, char *n, int up),
665 struct link *l_ptr, int up)
667 struct link_event *ev;
669 ev = kmalloc(sizeof(*ev), GFP_ATOMIC);
670 if (!ev) {
671 warn("Link event allocation failure\n");
672 return;
674 ev->addr = l_ptr->addr;
675 ev->up = up;
676 ev->fcn = fcn;
677 memcpy(ev->name, l_ptr->name, TIPC_MAX_LINK_NAME);
678 tipc_k_signal((Handler)link_recv_event, (unsigned long)ev);
681 #else
683 #define link_send_event(fcn, l_ptr, up) do { } while (0)
685 #endif
687 void tipc_link_reset(struct link *l_ptr)
689 struct sk_buff *buf;
690 u32 prev_state = l_ptr->state;
691 u32 checkpoint = l_ptr->next_in_no;
692 int was_active_link = tipc_link_is_active(l_ptr);
694 msg_set_session(l_ptr->pmsg, msg_session(l_ptr->pmsg) + 1);
696 /* Link is down, accept any session: */
697 l_ptr->peer_session = 0;
699 /* Prepare for max packet size negotiation */
700 link_init_max_pkt(l_ptr);
702 l_ptr->state = RESET_UNKNOWN;
703 dbg_link_state("Resetting Link\n");
705 if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
706 return;
708 tipc_node_link_down(l_ptr->owner, l_ptr);
709 tipc_bearer_remove_dest(l_ptr->b_ptr, l_ptr->addr);
710 #if 0
711 tipc_printf(TIPC_CONS, "\nReset link <%s>\n", l_ptr->name);
712 dbg_link_dump();
713 #endif
714 if (was_active_link && tipc_node_has_active_links(l_ptr->owner) &&
715 l_ptr->owner->permit_changeover) {
716 l_ptr->reset_checkpoint = checkpoint;
717 l_ptr->exp_msg_count = START_CHANGEOVER;
720 /* Clean up all queues: */
722 link_release_outqueue(l_ptr);
723 buf_discard(l_ptr->proto_msg_queue);
724 l_ptr->proto_msg_queue = NULL;
725 buf = l_ptr->oldest_deferred_in;
726 while (buf) {
727 struct sk_buff *next = buf->next;
728 buf_discard(buf);
729 buf = next;
731 if (!list_empty(&l_ptr->waiting_ports))
732 tipc_link_wakeup_ports(l_ptr, 1);
734 l_ptr->retransm_queue_head = 0;
735 l_ptr->retransm_queue_size = 0;
736 l_ptr->last_out = NULL;
737 l_ptr->first_out = NULL;
738 l_ptr->next_out = NULL;
739 l_ptr->unacked_window = 0;
740 l_ptr->checkpoint = 1;
741 l_ptr->next_out_no = 1;
742 l_ptr->deferred_inqueue_sz = 0;
743 l_ptr->oldest_deferred_in = NULL;
744 l_ptr->newest_deferred_in = NULL;
745 l_ptr->fsm_msg_cnt = 0;
746 l_ptr->stale_count = 0;
747 link_reset_statistics(l_ptr);
749 link_send_event(tipc_cfg_link_event, l_ptr, 0);
750 if (!in_own_cluster(l_ptr->addr))
751 link_send_event(tipc_disc_link_event, l_ptr, 0);
755 static void link_activate(struct link *l_ptr)
757 l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
758 tipc_node_link_up(l_ptr->owner, l_ptr);
759 tipc_bearer_add_dest(l_ptr->b_ptr, l_ptr->addr);
760 link_send_event(tipc_cfg_link_event, l_ptr, 1);
761 if (!in_own_cluster(l_ptr->addr))
762 link_send_event(tipc_disc_link_event, l_ptr, 1);
766 * link_state_event - link finite state machine
767 * @l_ptr: pointer to link
768 * @event: state machine event to process
771 static void link_state_event(struct link *l_ptr, unsigned event)
773 struct link *other;
774 u32 cont_intv = l_ptr->continuity_interval;
776 if (!l_ptr->started && (event != STARTING_EVT))
777 return; /* Not yet. */
779 if (link_blocked(l_ptr)) {
780 if (event == TIMEOUT_EVT) {
781 link_set_timer(l_ptr, cont_intv);
783 return; /* Changeover going on */
785 dbg_link("STATE_EV: <%s> ", l_ptr->name);
787 switch (l_ptr->state) {
788 case WORKING_WORKING:
789 dbg_link("WW/");
790 switch (event) {
791 case TRAFFIC_MSG_EVT:
792 dbg_link("TRF-");
793 /* fall through */
794 case ACTIVATE_MSG:
795 dbg_link("ACT\n");
796 break;
797 case TIMEOUT_EVT:
798 dbg_link("TIM ");
799 if (l_ptr->next_in_no != l_ptr->checkpoint) {
800 l_ptr->checkpoint = l_ptr->next_in_no;
801 if (tipc_bclink_acks_missing(l_ptr->owner)) {
802 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
803 0, 0, 0, 0, 0);
804 l_ptr->fsm_msg_cnt++;
805 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
806 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
807 1, 0, 0, 0, 0);
808 l_ptr->fsm_msg_cnt++;
810 link_set_timer(l_ptr, cont_intv);
811 break;
813 dbg_link(" -> WU\n");
814 l_ptr->state = WORKING_UNKNOWN;
815 l_ptr->fsm_msg_cnt = 0;
816 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
817 l_ptr->fsm_msg_cnt++;
818 link_set_timer(l_ptr, cont_intv / 4);
819 break;
820 case RESET_MSG:
821 dbg_link("RES -> RR\n");
822 info("Resetting link <%s>, requested by peer\n",
823 l_ptr->name);
824 tipc_link_reset(l_ptr);
825 l_ptr->state = RESET_RESET;
826 l_ptr->fsm_msg_cnt = 0;
827 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
828 l_ptr->fsm_msg_cnt++;
829 link_set_timer(l_ptr, cont_intv);
830 break;
831 default:
832 err("Unknown link event %u in WW state\n", event);
834 break;
835 case WORKING_UNKNOWN:
836 dbg_link("WU/");
837 switch (event) {
838 case TRAFFIC_MSG_EVT:
839 dbg_link("TRF-");
840 case ACTIVATE_MSG:
841 dbg_link("ACT -> WW\n");
842 l_ptr->state = WORKING_WORKING;
843 l_ptr->fsm_msg_cnt = 0;
844 link_set_timer(l_ptr, cont_intv);
845 break;
846 case RESET_MSG:
847 dbg_link("RES -> RR\n");
848 info("Resetting link <%s>, requested by peer "
849 "while probing\n", l_ptr->name);
850 tipc_link_reset(l_ptr);
851 l_ptr->state = RESET_RESET;
852 l_ptr->fsm_msg_cnt = 0;
853 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
854 l_ptr->fsm_msg_cnt++;
855 link_set_timer(l_ptr, cont_intv);
856 break;
857 case TIMEOUT_EVT:
858 dbg_link("TIM ");
859 if (l_ptr->next_in_no != l_ptr->checkpoint) {
860 dbg_link("-> WW \n");
861 l_ptr->state = WORKING_WORKING;
862 l_ptr->fsm_msg_cnt = 0;
863 l_ptr->checkpoint = l_ptr->next_in_no;
864 if (tipc_bclink_acks_missing(l_ptr->owner)) {
865 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
866 0, 0, 0, 0, 0);
867 l_ptr->fsm_msg_cnt++;
869 link_set_timer(l_ptr, cont_intv);
870 } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
871 dbg_link("Probing %u/%u,timer = %u ms)\n",
872 l_ptr->fsm_msg_cnt, l_ptr->abort_limit,
873 cont_intv / 4);
874 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
875 1, 0, 0, 0, 0);
876 l_ptr->fsm_msg_cnt++;
877 link_set_timer(l_ptr, cont_intv / 4);
878 } else { /* Link has failed */
879 dbg_link("-> RU (%u probes unanswered)\n",
880 l_ptr->fsm_msg_cnt);
881 warn("Resetting link <%s>, peer not responding\n",
882 l_ptr->name);
883 tipc_link_reset(l_ptr);
884 l_ptr->state = RESET_UNKNOWN;
885 l_ptr->fsm_msg_cnt = 0;
886 tipc_link_send_proto_msg(l_ptr, RESET_MSG,
887 0, 0, 0, 0, 0);
888 l_ptr->fsm_msg_cnt++;
889 link_set_timer(l_ptr, cont_intv);
891 break;
892 default:
893 err("Unknown link event %u in WU state\n", event);
895 break;
896 case RESET_UNKNOWN:
897 dbg_link("RU/");
898 switch (event) {
899 case TRAFFIC_MSG_EVT:
900 dbg_link("TRF-\n");
901 break;
902 case ACTIVATE_MSG:
903 other = l_ptr->owner->active_links[0];
904 if (other && link_working_unknown(other)) {
905 dbg_link("ACT\n");
906 break;
908 dbg_link("ACT -> WW\n");
909 l_ptr->state = WORKING_WORKING;
910 l_ptr->fsm_msg_cnt = 0;
911 link_activate(l_ptr);
912 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
913 l_ptr->fsm_msg_cnt++;
914 link_set_timer(l_ptr, cont_intv);
915 break;
916 case RESET_MSG:
917 dbg_link("RES \n");
918 dbg_link(" -> RR\n");
919 l_ptr->state = RESET_RESET;
920 l_ptr->fsm_msg_cnt = 0;
921 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 1, 0, 0, 0, 0);
922 l_ptr->fsm_msg_cnt++;
923 link_set_timer(l_ptr, cont_intv);
924 break;
925 case STARTING_EVT:
926 dbg_link("START-");
927 l_ptr->started = 1;
928 /* fall through */
929 case TIMEOUT_EVT:
930 dbg_link("TIM \n");
931 tipc_link_send_proto_msg(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
932 l_ptr->fsm_msg_cnt++;
933 link_set_timer(l_ptr, cont_intv);
934 break;
935 default:
936 err("Unknown link event %u in RU state\n", event);
938 break;
939 case RESET_RESET:
940 dbg_link("RR/ ");
941 switch (event) {
942 case TRAFFIC_MSG_EVT:
943 dbg_link("TRF-");
944 /* fall through */
945 case ACTIVATE_MSG:
946 other = l_ptr->owner->active_links[0];
947 if (other && link_working_unknown(other)) {
948 dbg_link("ACT\n");
949 break;
951 dbg_link("ACT -> WW\n");
952 l_ptr->state = WORKING_WORKING;
953 l_ptr->fsm_msg_cnt = 0;
954 link_activate(l_ptr);
955 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
956 l_ptr->fsm_msg_cnt++;
957 link_set_timer(l_ptr, cont_intv);
958 break;
959 case RESET_MSG:
960 dbg_link("RES\n");
961 break;
962 case TIMEOUT_EVT:
963 dbg_link("TIM\n");
964 tipc_link_send_proto_msg(l_ptr, ACTIVATE_MSG, 0, 0, 0, 0, 0);
965 l_ptr->fsm_msg_cnt++;
966 link_set_timer(l_ptr, cont_intv);
967 dbg_link("fsm_msg_cnt %u\n", l_ptr->fsm_msg_cnt);
968 break;
969 default:
970 err("Unknown link event %u in RR state\n", event);
972 break;
973 default:
974 err("Unknown link state %u/%u\n", l_ptr->state, event);
979 * link_bundle_buf(): Append contents of a buffer to
980 * the tail of an existing one.
983 static int link_bundle_buf(struct link *l_ptr,
984 struct sk_buff *bundler,
985 struct sk_buff *buf)
987 struct tipc_msg *bundler_msg = buf_msg(bundler);
988 struct tipc_msg *msg = buf_msg(buf);
989 u32 size = msg_size(msg);
990 u32 bundle_size = msg_size(bundler_msg);
991 u32 to_pos = align(bundle_size);
992 u32 pad = to_pos - bundle_size;
994 if (msg_user(bundler_msg) != MSG_BUNDLER)
995 return 0;
996 if (msg_type(bundler_msg) != OPEN_MSG)
997 return 0;
998 if (skb_tailroom(bundler) < (pad + size))
999 return 0;
1000 if (link_max_pkt(l_ptr) < (to_pos + size))
1001 return 0;
1003 skb_put(bundler, pad + size);
1004 memcpy(bundler->data + to_pos, buf->data, size);
1005 msg_set_size(bundler_msg, to_pos + size);
1006 msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
1007 dbg("Packed msg # %u(%u octets) into pos %u in buf(#%u)\n",
1008 msg_msgcnt(bundler_msg), size, to_pos, msg_seqno(bundler_msg));
1009 msg_dbg(msg, "PACKD:");
1010 buf_discard(buf);
1011 l_ptr->stats.sent_bundled++;
1012 return 1;
1015 static void link_add_to_outqueue(struct link *l_ptr,
1016 struct sk_buff *buf,
1017 struct tipc_msg *msg)
1019 u32 ack = mod(l_ptr->next_in_no - 1);
1020 u32 seqno = mod(l_ptr->next_out_no++);
1022 msg_set_word(msg, 2, ((ack << 16) | seqno));
1023 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1024 buf->next = NULL;
1025 if (l_ptr->first_out) {
1026 l_ptr->last_out->next = buf;
1027 l_ptr->last_out = buf;
1028 } else
1029 l_ptr->first_out = l_ptr->last_out = buf;
1030 l_ptr->out_queue_size++;
1034 * tipc_link_send_buf() is the 'full path' for messages, called from
1035 * inside TIPC when the 'fast path' in tipc_send_buf
1036 * has failed, and from link_send()
1039 int tipc_link_send_buf(struct link *l_ptr, struct sk_buff *buf)
1041 struct tipc_msg *msg = buf_msg(buf);
1042 u32 size = msg_size(msg);
1043 u32 dsz = msg_data_sz(msg);
1044 u32 queue_size = l_ptr->out_queue_size;
1045 u32 imp = msg_tot_importance(msg);
1046 u32 queue_limit = l_ptr->queue_limit[imp];
1047 u32 max_packet = link_max_pkt(l_ptr);
1049 msg_set_prevnode(msg, tipc_own_addr); /* If routed message */
1051 /* Match msg importance against queue limits: */
1053 if (unlikely(queue_size >= queue_limit)) {
1054 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
1055 return link_schedule_port(l_ptr, msg_origport(msg),
1056 size);
1058 msg_dbg(msg, "TIPC: Congestion, throwing away\n");
1059 buf_discard(buf);
1060 if (imp > CONN_MANAGER) {
1061 warn("Resetting link <%s>, send queue full", l_ptr->name);
1062 tipc_link_reset(l_ptr);
1064 return dsz;
1067 /* Fragmentation needed ? */
1069 if (size > max_packet)
1070 return tipc_link_send_long_buf(l_ptr, buf);
1072 /* Packet can be queued or sent: */
1074 if (queue_size > l_ptr->stats.max_queue_sz)
1075 l_ptr->stats.max_queue_sz = queue_size;
1077 if (likely(!tipc_bearer_congested(l_ptr->b_ptr, l_ptr) &&
1078 !link_congested(l_ptr))) {
1079 link_add_to_outqueue(l_ptr, buf, msg);
1081 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr))) {
1082 l_ptr->unacked_window = 0;
1083 } else {
1084 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1085 l_ptr->stats.bearer_congs++;
1086 l_ptr->next_out = buf;
1088 return dsz;
1090 /* Congestion: can message be bundled ?: */
1092 if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
1093 (msg_user(msg) != MSG_FRAGMENTER)) {
1095 /* Try adding message to an existing bundle */
1097 if (l_ptr->next_out &&
1098 link_bundle_buf(l_ptr, l_ptr->last_out, buf)) {
1099 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1100 return dsz;
1103 /* Try creating a new bundle */
1105 if (size <= max_packet * 2 / 3) {
1106 struct sk_buff *bundler = buf_acquire(max_packet);
1107 struct tipc_msg bundler_hdr;
1109 if (bundler) {
1110 msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
1111 TIPC_OK, INT_H_SIZE, l_ptr->addr);
1112 memcpy(bundler->data, (unchar *)&bundler_hdr,
1113 INT_H_SIZE);
1114 skb_trim(bundler, INT_H_SIZE);
1115 link_bundle_buf(l_ptr, bundler, buf);
1116 buf = bundler;
1117 msg = buf_msg(buf);
1118 l_ptr->stats.sent_bundles++;
1122 if (!l_ptr->next_out)
1123 l_ptr->next_out = buf;
1124 link_add_to_outqueue(l_ptr, buf, msg);
1125 tipc_bearer_resolve_congestion(l_ptr->b_ptr, l_ptr);
1126 return dsz;
1130 * tipc_link_send(): same as tipc_link_send_buf(), but the link to use has
1131 * not been selected yet, and the the owner node is not locked
1132 * Called by TIPC internal users, e.g. the name distributor
1135 int tipc_link_send(struct sk_buff *buf, u32 dest, u32 selector)
1137 struct link *l_ptr;
1138 struct node *n_ptr;
1139 int res = -ELINKCONG;
1141 read_lock_bh(&tipc_net_lock);
1142 n_ptr = tipc_node_select(dest, selector);
1143 if (n_ptr) {
1144 tipc_node_lock(n_ptr);
1145 l_ptr = n_ptr->active_links[selector & 1];
1146 if (l_ptr) {
1147 dbg("tipc_link_send: found link %x for dest %x\n", l_ptr, dest);
1148 res = tipc_link_send_buf(l_ptr, buf);
1149 } else {
1150 dbg("Attempt to send msg to unreachable node:\n");
1151 msg_dbg(buf_msg(buf),">>>");
1152 buf_discard(buf);
1154 tipc_node_unlock(n_ptr);
1155 } else {
1156 dbg("Attempt to send msg to unknown node:\n");
1157 msg_dbg(buf_msg(buf),">>>");
1158 buf_discard(buf);
1160 read_unlock_bh(&tipc_net_lock);
1161 return res;
1165 * link_send_buf_fast: Entry for data messages where the
1166 * destination link is known and the header is complete,
1167 * inclusive total message length. Very time critical.
1168 * Link is locked. Returns user data length.
1171 static int link_send_buf_fast(struct link *l_ptr, struct sk_buff *buf,
1172 u32 *used_max_pkt)
1174 struct tipc_msg *msg = buf_msg(buf);
1175 int res = msg_data_sz(msg);
1177 if (likely(!link_congested(l_ptr))) {
1178 if (likely(msg_size(msg) <= link_max_pkt(l_ptr))) {
1179 if (likely(list_empty(&l_ptr->b_ptr->cong_links))) {
1180 link_add_to_outqueue(l_ptr, buf, msg);
1181 if (likely(tipc_bearer_send(l_ptr->b_ptr, buf,
1182 &l_ptr->media_addr))) {
1183 l_ptr->unacked_window = 0;
1184 msg_dbg(msg,"SENT_FAST:");
1185 return res;
1187 dbg("failed sent fast...\n");
1188 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1189 l_ptr->stats.bearer_congs++;
1190 l_ptr->next_out = buf;
1191 return res;
1194 else
1195 *used_max_pkt = link_max_pkt(l_ptr);
1197 return tipc_link_send_buf(l_ptr, buf); /* All other cases */
1201 * tipc_send_buf_fast: Entry for data messages where the
1202 * destination node is known and the header is complete,
1203 * inclusive total message length.
1204 * Returns user data length.
1206 int tipc_send_buf_fast(struct sk_buff *buf, u32 destnode)
1208 struct link *l_ptr;
1209 struct node *n_ptr;
1210 int res;
1211 u32 selector = msg_origport(buf_msg(buf)) & 1;
1212 u32 dummy;
1214 if (destnode == tipc_own_addr)
1215 return tipc_port_recv_msg(buf);
1217 read_lock_bh(&tipc_net_lock);
1218 n_ptr = tipc_node_select(destnode, selector);
1219 if (likely(n_ptr)) {
1220 tipc_node_lock(n_ptr);
1221 l_ptr = n_ptr->active_links[selector];
1222 dbg("send_fast: buf %x selected %x, destnode = %x\n",
1223 buf, l_ptr, destnode);
1224 if (likely(l_ptr)) {
1225 res = link_send_buf_fast(l_ptr, buf, &dummy);
1226 tipc_node_unlock(n_ptr);
1227 read_unlock_bh(&tipc_net_lock);
1228 return res;
1230 tipc_node_unlock(n_ptr);
1232 read_unlock_bh(&tipc_net_lock);
1233 res = msg_data_sz(buf_msg(buf));
1234 tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1235 return res;
1240 * tipc_link_send_sections_fast: Entry for messages where the
1241 * destination processor is known and the header is complete,
1242 * except for total message length.
1243 * Returns user data length or errno.
1245 int tipc_link_send_sections_fast(struct port *sender,
1246 struct iovec const *msg_sect,
1247 const u32 num_sect,
1248 u32 destaddr)
1250 struct tipc_msg *hdr = &sender->publ.phdr;
1251 struct link *l_ptr;
1252 struct sk_buff *buf;
1253 struct node *node;
1254 int res;
1255 u32 selector = msg_origport(hdr) & 1;
1257 again:
1259 * Try building message using port's max_pkt hint.
1260 * (Must not hold any locks while building message.)
1263 res = msg_build(hdr, msg_sect, num_sect, sender->max_pkt,
1264 !sender->user_port, &buf);
1266 read_lock_bh(&tipc_net_lock);
1267 node = tipc_node_select(destaddr, selector);
1268 if (likely(node)) {
1269 tipc_node_lock(node);
1270 l_ptr = node->active_links[selector];
1271 if (likely(l_ptr)) {
1272 if (likely(buf)) {
1273 res = link_send_buf_fast(l_ptr, buf,
1274 &sender->max_pkt);
1275 if (unlikely(res < 0))
1276 buf_discard(buf);
1277 exit:
1278 tipc_node_unlock(node);
1279 read_unlock_bh(&tipc_net_lock);
1280 return res;
1283 /* Exit if build request was invalid */
1285 if (unlikely(res < 0))
1286 goto exit;
1288 /* Exit if link (or bearer) is congested */
1290 if (link_congested(l_ptr) ||
1291 !list_empty(&l_ptr->b_ptr->cong_links)) {
1292 res = link_schedule_port(l_ptr,
1293 sender->publ.ref, res);
1294 goto exit;
1298 * Message size exceeds max_pkt hint; update hint,
1299 * then re-try fast path or fragment the message
1302 sender->max_pkt = link_max_pkt(l_ptr);
1303 tipc_node_unlock(node);
1304 read_unlock_bh(&tipc_net_lock);
1307 if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1308 goto again;
1310 return link_send_sections_long(sender, msg_sect,
1311 num_sect, destaddr);
1313 tipc_node_unlock(node);
1315 read_unlock_bh(&tipc_net_lock);
1317 /* Couldn't find a link to the destination node */
1319 if (buf)
1320 return tipc_reject_msg(buf, TIPC_ERR_NO_NODE);
1321 if (res >= 0)
1322 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1323 TIPC_ERR_NO_NODE);
1324 return res;
1328 * link_send_sections_long(): Entry for long messages where the
1329 * destination node is known and the header is complete,
1330 * inclusive total message length.
1331 * Link and bearer congestion status have been checked to be ok,
1332 * and are ignored if they change.
1334 * Note that fragments do not use the full link MTU so that they won't have
1335 * to undergo refragmentation if link changeover causes them to be sent
1336 * over another link with an additional tunnel header added as prefix.
1337 * (Refragmentation will still occur if the other link has a smaller MTU.)
1339 * Returns user data length or errno.
1341 static int link_send_sections_long(struct port *sender,
1342 struct iovec const *msg_sect,
1343 u32 num_sect,
1344 u32 destaddr)
1346 struct link *l_ptr;
1347 struct node *node;
1348 struct tipc_msg *hdr = &sender->publ.phdr;
1349 u32 dsz = msg_data_sz(hdr);
1350 u32 max_pkt,fragm_sz,rest;
1351 struct tipc_msg fragm_hdr;
1352 struct sk_buff *buf,*buf_chain,*prev;
1353 u32 fragm_crs,fragm_rest,hsz,sect_rest;
1354 const unchar *sect_crs;
1355 int curr_sect;
1356 u32 fragm_no;
1358 again:
1359 fragm_no = 1;
1360 max_pkt = sender->max_pkt - INT_H_SIZE;
1361 /* leave room for tunnel header in case of link changeover */
1362 fragm_sz = max_pkt - INT_H_SIZE;
1363 /* leave room for fragmentation header in each fragment */
1364 rest = dsz;
1365 fragm_crs = 0;
1366 fragm_rest = 0;
1367 sect_rest = 0;
1368 sect_crs = NULL;
1369 curr_sect = -1;
1371 /* Prepare reusable fragment header: */
1373 msg_dbg(hdr, ">FRAGMENTING>");
1374 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1375 TIPC_OK, INT_H_SIZE, msg_destnode(hdr));
1376 msg_set_link_selector(&fragm_hdr, sender->publ.ref);
1377 msg_set_size(&fragm_hdr, max_pkt);
1378 msg_set_fragm_no(&fragm_hdr, 1);
1380 /* Prepare header of first fragment: */
1382 buf_chain = buf = buf_acquire(max_pkt);
1383 if (!buf)
1384 return -ENOMEM;
1385 buf->next = NULL;
1386 memcpy(buf->data, (unchar *)&fragm_hdr, INT_H_SIZE);
1387 hsz = msg_hdr_sz(hdr);
1388 memcpy(buf->data + INT_H_SIZE, (unchar *)hdr, hsz);
1389 msg_dbg(buf_msg(buf), ">BUILD>");
1391 /* Chop up message: */
1393 fragm_crs = INT_H_SIZE + hsz;
1394 fragm_rest = fragm_sz - hsz;
1396 do { /* For all sections */
1397 u32 sz;
1399 if (!sect_rest) {
1400 sect_rest = msg_sect[++curr_sect].iov_len;
1401 sect_crs = (const unchar *)msg_sect[curr_sect].iov_base;
1404 if (sect_rest < fragm_rest)
1405 sz = sect_rest;
1406 else
1407 sz = fragm_rest;
1409 if (likely(!sender->user_port)) {
1410 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1411 error:
1412 for (; buf_chain; buf_chain = buf) {
1413 buf = buf_chain->next;
1414 buf_discard(buf_chain);
1416 return -EFAULT;
1418 } else
1419 memcpy(buf->data + fragm_crs, sect_crs, sz);
1421 sect_crs += sz;
1422 sect_rest -= sz;
1423 fragm_crs += sz;
1424 fragm_rest -= sz;
1425 rest -= sz;
1427 if (!fragm_rest && rest) {
1429 /* Initiate new fragment: */
1430 if (rest <= fragm_sz) {
1431 fragm_sz = rest;
1432 msg_set_type(&fragm_hdr,LAST_FRAGMENT);
1433 } else {
1434 msg_set_type(&fragm_hdr, FRAGMENT);
1436 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1437 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1438 prev = buf;
1439 buf = buf_acquire(fragm_sz + INT_H_SIZE);
1440 if (!buf)
1441 goto error;
1443 buf->next = NULL;
1444 prev->next = buf;
1445 memcpy(buf->data, (unchar *)&fragm_hdr, INT_H_SIZE);
1446 fragm_crs = INT_H_SIZE;
1447 fragm_rest = fragm_sz;
1448 msg_dbg(buf_msg(buf)," >BUILD>");
1451 while (rest > 0);
1454 * Now we have a buffer chain. Select a link and check
1455 * that packet size is still OK
1457 node = tipc_node_select(destaddr, sender->publ.ref & 1);
1458 if (likely(node)) {
1459 tipc_node_lock(node);
1460 l_ptr = node->active_links[sender->publ.ref & 1];
1461 if (!l_ptr) {
1462 tipc_node_unlock(node);
1463 goto reject;
1465 if (link_max_pkt(l_ptr) < max_pkt) {
1466 sender->max_pkt = link_max_pkt(l_ptr);
1467 tipc_node_unlock(node);
1468 for (; buf_chain; buf_chain = buf) {
1469 buf = buf_chain->next;
1470 buf_discard(buf_chain);
1472 goto again;
1474 } else {
1475 reject:
1476 for (; buf_chain; buf_chain = buf) {
1477 buf = buf_chain->next;
1478 buf_discard(buf_chain);
1480 return tipc_port_reject_sections(sender, hdr, msg_sect, num_sect,
1481 TIPC_ERR_NO_NODE);
1484 /* Append whole chain to send queue: */
1486 buf = buf_chain;
1487 l_ptr->long_msg_seq_no = mod(l_ptr->long_msg_seq_no + 1);
1488 if (!l_ptr->next_out)
1489 l_ptr->next_out = buf_chain;
1490 l_ptr->stats.sent_fragmented++;
1491 while (buf) {
1492 struct sk_buff *next = buf->next;
1493 struct tipc_msg *msg = buf_msg(buf);
1495 l_ptr->stats.sent_fragments++;
1496 msg_set_long_msgno(msg, l_ptr->long_msg_seq_no);
1497 link_add_to_outqueue(l_ptr, buf, msg);
1498 msg_dbg(msg, ">ADD>");
1499 buf = next;
1502 /* Send it, if possible: */
1504 tipc_link_push_queue(l_ptr);
1505 tipc_node_unlock(node);
1506 return dsz;
1510 * tipc_link_push_packet: Push one unsent packet to the media
1512 u32 tipc_link_push_packet(struct link *l_ptr)
1514 struct sk_buff *buf = l_ptr->first_out;
1515 u32 r_q_size = l_ptr->retransm_queue_size;
1516 u32 r_q_head = l_ptr->retransm_queue_head;
1518 /* Step to position where retransmission failed, if any, */
1519 /* consider that buffers may have been released in meantime */
1521 if (r_q_size && buf) {
1522 u32 last = lesser(mod(r_q_head + r_q_size),
1523 link_last_sent(l_ptr));
1524 u32 first = msg_seqno(buf_msg(buf));
1526 while (buf && less(first, r_q_head)) {
1527 first = mod(first + 1);
1528 buf = buf->next;
1530 l_ptr->retransm_queue_head = r_q_head = first;
1531 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1534 /* Continue retransmission now, if there is anything: */
1536 if (r_q_size && buf && !skb_cloned(buf)) {
1537 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1538 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1539 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1540 msg_dbg(buf_msg(buf), ">DEF-RETR>");
1541 l_ptr->retransm_queue_head = mod(++r_q_head);
1542 l_ptr->retransm_queue_size = --r_q_size;
1543 l_ptr->stats.retransmitted++;
1544 return TIPC_OK;
1545 } else {
1546 l_ptr->stats.bearer_congs++;
1547 msg_dbg(buf_msg(buf), "|>DEF-RETR>");
1548 return PUSH_FAILED;
1552 /* Send deferred protocol message, if any: */
1554 buf = l_ptr->proto_msg_queue;
1555 if (buf) {
1556 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1557 msg_set_bcast_ack(buf_msg(buf),l_ptr->owner->bclink.last_in);
1558 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1559 msg_dbg(buf_msg(buf), ">DEF-PROT>");
1560 l_ptr->unacked_window = 0;
1561 buf_discard(buf);
1562 l_ptr->proto_msg_queue = NULL;
1563 return TIPC_OK;
1564 } else {
1565 msg_dbg(buf_msg(buf), "|>DEF-PROT>");
1566 l_ptr->stats.bearer_congs++;
1567 return PUSH_FAILED;
1571 /* Send one deferred data message, if send window not full: */
1573 buf = l_ptr->next_out;
1574 if (buf) {
1575 struct tipc_msg *msg = buf_msg(buf);
1576 u32 next = msg_seqno(msg);
1577 u32 first = msg_seqno(buf_msg(l_ptr->first_out));
1579 if (mod(next - first) < l_ptr->queue_limit[0]) {
1580 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1581 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1582 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1583 if (msg_user(msg) == MSG_BUNDLER)
1584 msg_set_type(msg, CLOSED_MSG);
1585 msg_dbg(msg, ">PUSH-DATA>");
1586 l_ptr->next_out = buf->next;
1587 return TIPC_OK;
1588 } else {
1589 msg_dbg(msg, "|PUSH-DATA|");
1590 l_ptr->stats.bearer_congs++;
1591 return PUSH_FAILED;
1595 return PUSH_FINISHED;
1599 * push_queue(): push out the unsent messages of a link where
1600 * congestion has abated. Node is locked
1602 void tipc_link_push_queue(struct link *l_ptr)
1604 u32 res;
1606 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr))
1607 return;
1609 do {
1610 res = tipc_link_push_packet(l_ptr);
1612 while (res == TIPC_OK);
1613 if (res == PUSH_FAILED)
1614 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1617 static void link_reset_all(unsigned long addr)
1619 struct node *n_ptr;
1620 char addr_string[16];
1621 u32 i;
1623 read_lock_bh(&tipc_net_lock);
1624 n_ptr = tipc_node_find((u32)addr);
1625 if (!n_ptr) {
1626 read_unlock_bh(&tipc_net_lock);
1627 return; /* node no longer exists */
1630 tipc_node_lock(n_ptr);
1632 warn("Resetting all links to %s\n",
1633 addr_string_fill(addr_string, n_ptr->addr));
1635 for (i = 0; i < MAX_BEARERS; i++) {
1636 if (n_ptr->links[i]) {
1637 link_print(n_ptr->links[i], TIPC_OUTPUT,
1638 "Resetting link\n");
1639 tipc_link_reset(n_ptr->links[i]);
1643 tipc_node_unlock(n_ptr);
1644 read_unlock_bh(&tipc_net_lock);
1647 static void link_retransmit_failure(struct link *l_ptr, struct sk_buff *buf)
1649 struct tipc_msg *msg = buf_msg(buf);
1651 warn("Retransmission failure on link <%s>\n", l_ptr->name);
1652 tipc_msg_print(TIPC_OUTPUT, msg, ">RETR-FAIL>");
1654 if (l_ptr->addr) {
1656 /* Handle failure on standard link */
1658 link_print(l_ptr, TIPC_OUTPUT, "Resetting link\n");
1659 tipc_link_reset(l_ptr);
1661 } else {
1663 /* Handle failure on broadcast link */
1665 struct node *n_ptr;
1666 char addr_string[16];
1668 tipc_printf(TIPC_OUTPUT, "Msg seq number: %u, ", msg_seqno(msg));
1669 tipc_printf(TIPC_OUTPUT, "Outstanding acks: %lu\n",
1670 (unsigned long) TIPC_SKB_CB(buf)->handle);
1672 n_ptr = l_ptr->owner->next;
1673 tipc_node_lock(n_ptr);
1675 addr_string_fill(addr_string, n_ptr->addr);
1676 tipc_printf(TIPC_OUTPUT, "Multicast link info for %s\n", addr_string);
1677 tipc_printf(TIPC_OUTPUT, "Supported: %d, ", n_ptr->bclink.supported);
1678 tipc_printf(TIPC_OUTPUT, "Acked: %u\n", n_ptr->bclink.acked);
1679 tipc_printf(TIPC_OUTPUT, "Last in: %u, ", n_ptr->bclink.last_in);
1680 tipc_printf(TIPC_OUTPUT, "Gap after: %u, ", n_ptr->bclink.gap_after);
1681 tipc_printf(TIPC_OUTPUT, "Gap to: %u\n", n_ptr->bclink.gap_to);
1682 tipc_printf(TIPC_OUTPUT, "Nack sync: %u\n\n", n_ptr->bclink.nack_sync);
1684 tipc_k_signal((Handler)link_reset_all, (unsigned long)n_ptr->addr);
1686 tipc_node_unlock(n_ptr);
1688 l_ptr->stale_count = 0;
1692 void tipc_link_retransmit(struct link *l_ptr, struct sk_buff *buf,
1693 u32 retransmits)
1695 struct tipc_msg *msg;
1697 if (!buf)
1698 return;
1700 msg = buf_msg(buf);
1702 dbg("Retransmitting %u in link %x\n", retransmits, l_ptr);
1704 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
1705 if (!skb_cloned(buf)) {
1706 msg_dbg(msg, ">NO_RETR->BCONG>");
1707 dbg_print_link(l_ptr, " ");
1708 l_ptr->retransm_queue_head = msg_seqno(msg);
1709 l_ptr->retransm_queue_size = retransmits;
1710 return;
1711 } else {
1712 /* Don't retransmit if driver already has the buffer */
1714 } else {
1715 /* Detect repeated retransmit failures on uncongested bearer */
1717 if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1718 if (++l_ptr->stale_count > 100) {
1719 link_retransmit_failure(l_ptr, buf);
1720 return;
1722 } else {
1723 l_ptr->last_retransmitted = msg_seqno(msg);
1724 l_ptr->stale_count = 1;
1728 while (retransmits && (buf != l_ptr->next_out) && buf && !skb_cloned(buf)) {
1729 msg = buf_msg(buf);
1730 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1731 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1732 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
1733 msg_dbg(buf_msg(buf), ">RETR>");
1734 buf = buf->next;
1735 retransmits--;
1736 l_ptr->stats.retransmitted++;
1737 } else {
1738 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
1739 l_ptr->stats.bearer_congs++;
1740 l_ptr->retransm_queue_head = msg_seqno(buf_msg(buf));
1741 l_ptr->retransm_queue_size = retransmits;
1742 return;
1746 l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1750 * link_recv_non_seq: Receive packets which are outside
1751 * the link sequence flow
1754 static void link_recv_non_seq(struct sk_buff *buf)
1756 struct tipc_msg *msg = buf_msg(buf);
1758 if (msg_user(msg) == LINK_CONFIG)
1759 tipc_disc_recv_msg(buf);
1760 else
1761 tipc_bclink_recv_pkt(buf);
1764 /**
1765 * link_insert_deferred_queue - insert deferred messages back into receive chain
1768 static struct sk_buff *link_insert_deferred_queue(struct link *l_ptr,
1769 struct sk_buff *buf)
1771 u32 seq_no;
1773 if (l_ptr->oldest_deferred_in == NULL)
1774 return buf;
1776 seq_no = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
1777 if (seq_no == mod(l_ptr->next_in_no)) {
1778 l_ptr->newest_deferred_in->next = buf;
1779 buf = l_ptr->oldest_deferred_in;
1780 l_ptr->oldest_deferred_in = NULL;
1781 l_ptr->deferred_inqueue_sz = 0;
1783 return buf;
1786 void tipc_recv_msg(struct sk_buff *head, struct tipc_bearer *tb_ptr)
1788 read_lock_bh(&tipc_net_lock);
1789 while (head) {
1790 struct bearer *b_ptr;
1791 struct node *n_ptr;
1792 struct link *l_ptr;
1793 struct sk_buff *crs;
1794 struct sk_buff *buf = head;
1795 struct tipc_msg *msg = buf_msg(buf);
1796 u32 seq_no = msg_seqno(msg);
1797 u32 ackd = msg_ack(msg);
1798 u32 released = 0;
1799 int type;
1801 b_ptr = (struct bearer *)tb_ptr;
1802 TIPC_SKB_CB(buf)->handle = b_ptr;
1804 head = head->next;
1805 if (unlikely(msg_version(msg) != TIPC_VERSION))
1806 goto cont;
1807 #if 0
1808 if (msg_user(msg) != LINK_PROTOCOL)
1809 #endif
1810 msg_dbg(msg,"<REC<");
1812 if (unlikely(msg_non_seq(msg))) {
1813 link_recv_non_seq(buf);
1814 continue;
1817 if (unlikely(!msg_short(msg) &&
1818 (msg_destnode(msg) != tipc_own_addr)))
1819 goto cont;
1821 n_ptr = tipc_node_find(msg_prevnode(msg));
1822 if (unlikely(!n_ptr))
1823 goto cont;
1825 tipc_node_lock(n_ptr);
1826 l_ptr = n_ptr->links[b_ptr->identity];
1827 if (unlikely(!l_ptr)) {
1828 tipc_node_unlock(n_ptr);
1829 goto cont;
1832 * Release acked messages
1834 if (less(n_ptr->bclink.acked, msg_bcast_ack(msg))) {
1835 if (tipc_node_is_up(n_ptr) && n_ptr->bclink.supported)
1836 tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1839 crs = l_ptr->first_out;
1840 while ((crs != l_ptr->next_out) &&
1841 less_eq(msg_seqno(buf_msg(crs)), ackd)) {
1842 struct sk_buff *next = crs->next;
1844 buf_discard(crs);
1845 crs = next;
1846 released++;
1848 if (released) {
1849 l_ptr->first_out = crs;
1850 l_ptr->out_queue_size -= released;
1852 if (unlikely(l_ptr->next_out))
1853 tipc_link_push_queue(l_ptr);
1854 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1855 tipc_link_wakeup_ports(l_ptr, 0);
1856 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1857 l_ptr->stats.sent_acks++;
1858 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1861 protocol_check:
1862 if (likely(link_working_working(l_ptr))) {
1863 if (likely(seq_no == mod(l_ptr->next_in_no))) {
1864 l_ptr->next_in_no++;
1865 if (unlikely(l_ptr->oldest_deferred_in))
1866 head = link_insert_deferred_queue(l_ptr,
1867 head);
1868 if (likely(msg_is_dest(msg, tipc_own_addr))) {
1869 deliver:
1870 if (likely(msg_isdata(msg))) {
1871 tipc_node_unlock(n_ptr);
1872 tipc_port_recv_msg(buf);
1873 continue;
1875 switch (msg_user(msg)) {
1876 case MSG_BUNDLER:
1877 l_ptr->stats.recv_bundles++;
1878 l_ptr->stats.recv_bundled +=
1879 msg_msgcnt(msg);
1880 tipc_node_unlock(n_ptr);
1881 tipc_link_recv_bundle(buf);
1882 continue;
1883 case ROUTE_DISTRIBUTOR:
1884 tipc_node_unlock(n_ptr);
1885 tipc_cltr_recv_routing_table(buf);
1886 continue;
1887 case NAME_DISTRIBUTOR:
1888 tipc_node_unlock(n_ptr);
1889 tipc_named_recv(buf);
1890 continue;
1891 case CONN_MANAGER:
1892 tipc_node_unlock(n_ptr);
1893 tipc_port_recv_proto_msg(buf);
1894 continue;
1895 case MSG_FRAGMENTER:
1896 l_ptr->stats.recv_fragments++;
1897 if (tipc_link_recv_fragment(&l_ptr->defragm_buf,
1898 &buf, &msg)) {
1899 l_ptr->stats.recv_fragmented++;
1900 goto deliver;
1902 break;
1903 case CHANGEOVER_PROTOCOL:
1904 type = msg_type(msg);
1905 if (link_recv_changeover_msg(&l_ptr, &buf)) {
1906 msg = buf_msg(buf);
1907 seq_no = msg_seqno(msg);
1908 TIPC_SKB_CB(buf)->handle
1909 = b_ptr;
1910 if (type == ORIGINAL_MSG)
1911 goto deliver;
1912 goto protocol_check;
1914 break;
1917 tipc_node_unlock(n_ptr);
1918 tipc_net_route_msg(buf);
1919 continue;
1921 link_handle_out_of_seq_msg(l_ptr, buf);
1922 head = link_insert_deferred_queue(l_ptr, head);
1923 tipc_node_unlock(n_ptr);
1924 continue;
1927 if (msg_user(msg) == LINK_PROTOCOL) {
1928 link_recv_proto_msg(l_ptr, buf);
1929 head = link_insert_deferred_queue(l_ptr, head);
1930 tipc_node_unlock(n_ptr);
1931 continue;
1933 msg_dbg(msg,"NSEQ<REC<");
1934 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1936 if (link_working_working(l_ptr)) {
1937 /* Re-insert in front of queue */
1938 msg_dbg(msg,"RECV-REINS:");
1939 buf->next = head;
1940 head = buf;
1941 tipc_node_unlock(n_ptr);
1942 continue;
1944 tipc_node_unlock(n_ptr);
1945 cont:
1946 buf_discard(buf);
1948 read_unlock_bh(&tipc_net_lock);
1952 * link_defer_buf(): Sort a received out-of-sequence packet
1953 * into the deferred reception queue.
1954 * Returns the increase of the queue length,i.e. 0 or 1
1957 u32 tipc_link_defer_pkt(struct sk_buff **head,
1958 struct sk_buff **tail,
1959 struct sk_buff *buf)
1961 struct sk_buff *prev = NULL;
1962 struct sk_buff *crs = *head;
1963 u32 seq_no = msg_seqno(buf_msg(buf));
1965 buf->next = NULL;
1967 /* Empty queue ? */
1968 if (*head == NULL) {
1969 *head = *tail = buf;
1970 return 1;
1973 /* Last ? */
1974 if (less(msg_seqno(buf_msg(*tail)), seq_no)) {
1975 (*tail)->next = buf;
1976 *tail = buf;
1977 return 1;
1980 /* Scan through queue and sort it in */
1981 do {
1982 struct tipc_msg *msg = buf_msg(crs);
1984 if (less(seq_no, msg_seqno(msg))) {
1985 buf->next = crs;
1986 if (prev)
1987 prev->next = buf;
1988 else
1989 *head = buf;
1990 return 1;
1992 if (seq_no == msg_seqno(msg)) {
1993 break;
1995 prev = crs;
1996 crs = crs->next;
1998 while (crs);
2000 /* Message is a duplicate of an existing message */
2002 buf_discard(buf);
2003 return 0;
2006 /**
2007 * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
2010 static void link_handle_out_of_seq_msg(struct link *l_ptr,
2011 struct sk_buff *buf)
2013 u32 seq_no = msg_seqno(buf_msg(buf));
2015 if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
2016 link_recv_proto_msg(l_ptr, buf);
2017 return;
2020 dbg("rx OOS msg: seq_no %u, expecting %u (%u)\n",
2021 seq_no, mod(l_ptr->next_in_no), l_ptr->next_in_no);
2023 /* Record OOS packet arrival (force mismatch on next timeout) */
2025 l_ptr->checkpoint--;
2028 * Discard packet if a duplicate; otherwise add it to deferred queue
2029 * and notify peer of gap as per protocol specification
2032 if (less(seq_no, mod(l_ptr->next_in_no))) {
2033 l_ptr->stats.duplicates++;
2034 buf_discard(buf);
2035 return;
2038 if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
2039 &l_ptr->newest_deferred_in, buf)) {
2040 l_ptr->deferred_inqueue_sz++;
2041 l_ptr->stats.deferred_recv++;
2042 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
2043 tipc_link_send_proto_msg(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
2044 } else
2045 l_ptr->stats.duplicates++;
2049 * Send protocol message to the other endpoint.
2051 void tipc_link_send_proto_msg(struct link *l_ptr, u32 msg_typ, int probe_msg,
2052 u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
2054 struct sk_buff *buf = NULL;
2055 struct tipc_msg *msg = l_ptr->pmsg;
2056 u32 msg_size = sizeof(l_ptr->proto_msg);
2058 if (link_blocked(l_ptr))
2059 return;
2060 msg_set_type(msg, msg_typ);
2061 msg_set_net_plane(msg, l_ptr->b_ptr->net_plane);
2062 msg_set_bcast_ack(msg, mod(l_ptr->owner->bclink.last_in));
2063 msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
2065 if (msg_typ == STATE_MSG) {
2066 u32 next_sent = mod(l_ptr->next_out_no);
2068 if (!tipc_link_is_up(l_ptr))
2069 return;
2070 if (l_ptr->next_out)
2071 next_sent = msg_seqno(buf_msg(l_ptr->next_out));
2072 msg_set_next_sent(msg, next_sent);
2073 if (l_ptr->oldest_deferred_in) {
2074 u32 rec = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
2075 gap = mod(rec - mod(l_ptr->next_in_no));
2077 msg_set_seq_gap(msg, gap);
2078 if (gap)
2079 l_ptr->stats.sent_nacks++;
2080 msg_set_link_tolerance(msg, tolerance);
2081 msg_set_linkprio(msg, priority);
2082 msg_set_max_pkt(msg, ack_mtu);
2083 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
2084 msg_set_probe(msg, probe_msg != 0);
2085 if (probe_msg) {
2086 u32 mtu = l_ptr->max_pkt;
2088 if ((mtu < l_ptr->max_pkt_target) &&
2089 link_working_working(l_ptr) &&
2090 l_ptr->fsm_msg_cnt) {
2091 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2092 if (l_ptr->max_pkt_probes == 10) {
2093 l_ptr->max_pkt_target = (msg_size - 4);
2094 l_ptr->max_pkt_probes = 0;
2095 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
2097 l_ptr->max_pkt_probes++;
2100 l_ptr->stats.sent_probes++;
2102 l_ptr->stats.sent_states++;
2103 } else { /* RESET_MSG or ACTIVATE_MSG */
2104 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
2105 msg_set_seq_gap(msg, 0);
2106 msg_set_next_sent(msg, 1);
2107 msg_set_link_tolerance(msg, l_ptr->tolerance);
2108 msg_set_linkprio(msg, l_ptr->priority);
2109 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
2112 if (tipc_node_has_redundant_links(l_ptr->owner)) {
2113 msg_set_redundant_link(msg);
2114 } else {
2115 msg_clear_redundant_link(msg);
2117 msg_set_linkprio(msg, l_ptr->priority);
2119 /* Ensure sequence number will not fit : */
2121 msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
2123 /* Congestion? */
2125 if (tipc_bearer_congested(l_ptr->b_ptr, l_ptr)) {
2126 if (!l_ptr->proto_msg_queue) {
2127 l_ptr->proto_msg_queue =
2128 buf_acquire(sizeof(l_ptr->proto_msg));
2130 buf = l_ptr->proto_msg_queue;
2131 if (!buf)
2132 return;
2133 memcpy(buf->data, (unchar *)msg, sizeof(l_ptr->proto_msg));
2134 return;
2136 msg_set_timestamp(msg, jiffies_to_msecs(jiffies));
2138 /* Message can be sent */
2140 msg_dbg(msg, ">>");
2142 buf = buf_acquire(msg_size);
2143 if (!buf)
2144 return;
2146 memcpy(buf->data, (unchar *)msg, sizeof(l_ptr->proto_msg));
2147 msg_set_size(buf_msg(buf), msg_size);
2149 if (tipc_bearer_send(l_ptr->b_ptr, buf, &l_ptr->media_addr)) {
2150 l_ptr->unacked_window = 0;
2151 buf_discard(buf);
2152 return;
2155 /* New congestion */
2156 tipc_bearer_schedule(l_ptr->b_ptr, l_ptr);
2157 l_ptr->proto_msg_queue = buf;
2158 l_ptr->stats.bearer_congs++;
2162 * Receive protocol message :
2163 * Note that network plane id propagates through the network, and may
2164 * change at any time. The node with lowest address rules
2167 static void link_recv_proto_msg(struct link *l_ptr, struct sk_buff *buf)
2169 u32 rec_gap = 0;
2170 u32 max_pkt_info;
2171 u32 max_pkt_ack;
2172 u32 msg_tol;
2173 struct tipc_msg *msg = buf_msg(buf);
2175 dbg("AT(%u):", jiffies_to_msecs(jiffies));
2176 msg_dbg(msg, "<<");
2177 if (link_blocked(l_ptr))
2178 goto exit;
2180 /* record unnumbered packet arrival (force mismatch on next timeout) */
2182 l_ptr->checkpoint--;
2184 if (l_ptr->b_ptr->net_plane != msg_net_plane(msg))
2185 if (tipc_own_addr > msg_prevnode(msg))
2186 l_ptr->b_ptr->net_plane = msg_net_plane(msg);
2188 l_ptr->owner->permit_changeover = msg_redundant_link(msg);
2190 switch (msg_type(msg)) {
2192 case RESET_MSG:
2193 if (!link_working_unknown(l_ptr) && l_ptr->peer_session) {
2194 if (msg_session(msg) == l_ptr->peer_session) {
2195 dbg("Duplicate RESET: %u<->%u\n",
2196 msg_session(msg), l_ptr->peer_session);
2197 break; /* duplicate: ignore */
2200 /* fall thru' */
2201 case ACTIVATE_MSG:
2202 /* Update link settings according other endpoint's values */
2204 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
2206 if ((msg_tol = msg_link_tolerance(msg)) &&
2207 (msg_tol > l_ptr->tolerance))
2208 link_set_supervision_props(l_ptr, msg_tol);
2210 if (msg_linkprio(msg) > l_ptr->priority)
2211 l_ptr->priority = msg_linkprio(msg);
2213 max_pkt_info = msg_max_pkt(msg);
2214 if (max_pkt_info) {
2215 if (max_pkt_info < l_ptr->max_pkt_target)
2216 l_ptr->max_pkt_target = max_pkt_info;
2217 if (l_ptr->max_pkt > l_ptr->max_pkt_target)
2218 l_ptr->max_pkt = l_ptr->max_pkt_target;
2219 } else {
2220 l_ptr->max_pkt = l_ptr->max_pkt_target;
2222 l_ptr->owner->bclink.supported = (max_pkt_info != 0);
2224 link_state_event(l_ptr, msg_type(msg));
2226 l_ptr->peer_session = msg_session(msg);
2227 l_ptr->peer_bearer_id = msg_bearer_id(msg);
2229 /* Synchronize broadcast sequence numbers */
2230 if (!tipc_node_has_redundant_links(l_ptr->owner)) {
2231 l_ptr->owner->bclink.last_in = mod(msg_last_bcast(msg));
2233 break;
2234 case STATE_MSG:
2236 if ((msg_tol = msg_link_tolerance(msg)))
2237 link_set_supervision_props(l_ptr, msg_tol);
2239 if (msg_linkprio(msg) &&
2240 (msg_linkprio(msg) != l_ptr->priority)) {
2241 warn("Resetting link <%s>, priority change %u->%u\n",
2242 l_ptr->name, l_ptr->priority, msg_linkprio(msg));
2243 l_ptr->priority = msg_linkprio(msg);
2244 tipc_link_reset(l_ptr); /* Enforce change to take effect */
2245 break;
2247 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
2248 l_ptr->stats.recv_states++;
2249 if (link_reset_unknown(l_ptr))
2250 break;
2252 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
2253 rec_gap = mod(msg_next_sent(msg) -
2254 mod(l_ptr->next_in_no));
2257 max_pkt_ack = msg_max_pkt(msg);
2258 if (max_pkt_ack > l_ptr->max_pkt) {
2259 dbg("Link <%s> updated MTU %u -> %u\n",
2260 l_ptr->name, l_ptr->max_pkt, max_pkt_ack);
2261 l_ptr->max_pkt = max_pkt_ack;
2262 l_ptr->max_pkt_probes = 0;
2265 max_pkt_ack = 0;
2266 if (msg_probe(msg)) {
2267 l_ptr->stats.recv_probes++;
2268 if (msg_size(msg) > sizeof(l_ptr->proto_msg)) {
2269 max_pkt_ack = msg_size(msg);
2273 /* Protocol message before retransmits, reduce loss risk */
2275 tipc_bclink_check_gap(l_ptr->owner, msg_last_bcast(msg));
2277 if (rec_gap || (msg_probe(msg))) {
2278 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2279 0, rec_gap, 0, 0, max_pkt_ack);
2281 if (msg_seq_gap(msg)) {
2282 msg_dbg(msg, "With Gap:");
2283 l_ptr->stats.recv_nacks++;
2284 tipc_link_retransmit(l_ptr, l_ptr->first_out,
2285 msg_seq_gap(msg));
2287 break;
2288 default:
2289 msg_dbg(buf_msg(buf), "<DISCARDING UNKNOWN<");
2291 exit:
2292 buf_discard(buf);
2297 * tipc_link_tunnel(): Send one message via a link belonging to
2298 * another bearer. Owner node is locked.
2300 void tipc_link_tunnel(struct link *l_ptr,
2301 struct tipc_msg *tunnel_hdr,
2302 struct tipc_msg *msg,
2303 u32 selector)
2305 struct link *tunnel;
2306 struct sk_buff *buf;
2307 u32 length = msg_size(msg);
2309 tunnel = l_ptr->owner->active_links[selector & 1];
2310 if (!tipc_link_is_up(tunnel)) {
2311 warn("Link changeover error, "
2312 "tunnel link no longer available\n");
2313 return;
2315 msg_set_size(tunnel_hdr, length + INT_H_SIZE);
2316 buf = buf_acquire(length + INT_H_SIZE);
2317 if (!buf) {
2318 warn("Link changeover error, "
2319 "unable to send tunnel msg\n");
2320 return;
2322 memcpy(buf->data, (unchar *)tunnel_hdr, INT_H_SIZE);
2323 memcpy(buf->data + INT_H_SIZE, (unchar *)msg, length);
2324 dbg("%c->%c:", l_ptr->b_ptr->net_plane, tunnel->b_ptr->net_plane);
2325 msg_dbg(buf_msg(buf), ">SEND>");
2326 tipc_link_send_buf(tunnel, buf);
2332 * changeover(): Send whole message queue via the remaining link
2333 * Owner node is locked.
2336 void tipc_link_changeover(struct link *l_ptr)
2338 u32 msgcount = l_ptr->out_queue_size;
2339 struct sk_buff *crs = l_ptr->first_out;
2340 struct link *tunnel = l_ptr->owner->active_links[0];
2341 struct tipc_msg tunnel_hdr;
2342 int split_bundles;
2344 if (!tunnel)
2345 return;
2347 if (!l_ptr->owner->permit_changeover) {
2348 warn("Link changeover error, "
2349 "peer did not permit changeover\n");
2350 return;
2353 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2354 ORIGINAL_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2355 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2356 msg_set_msgcnt(&tunnel_hdr, msgcount);
2357 dbg("Link changeover requires %u tunnel messages\n", msgcount);
2359 if (!l_ptr->first_out) {
2360 struct sk_buff *buf;
2362 buf = buf_acquire(INT_H_SIZE);
2363 if (buf) {
2364 memcpy(buf->data, (unchar *)&tunnel_hdr, INT_H_SIZE);
2365 msg_set_size(&tunnel_hdr, INT_H_SIZE);
2366 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2367 tunnel->b_ptr->net_plane);
2368 msg_dbg(&tunnel_hdr, "EMPTY>SEND>");
2369 tipc_link_send_buf(tunnel, buf);
2370 } else {
2371 warn("Link changeover error, "
2372 "unable to send changeover msg\n");
2374 return;
2377 split_bundles = (l_ptr->owner->active_links[0] !=
2378 l_ptr->owner->active_links[1]);
2380 while (crs) {
2381 struct tipc_msg *msg = buf_msg(crs);
2383 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2384 u32 msgcount = msg_msgcnt(msg);
2385 struct tipc_msg *m = msg_get_wrapped(msg);
2386 unchar* pos = (unchar*)m;
2388 while (msgcount--) {
2389 msg_set_seqno(m,msg_seqno(msg));
2390 tipc_link_tunnel(l_ptr, &tunnel_hdr, m,
2391 msg_link_selector(m));
2392 pos += align(msg_size(m));
2393 m = (struct tipc_msg *)pos;
2395 } else {
2396 tipc_link_tunnel(l_ptr, &tunnel_hdr, msg,
2397 msg_link_selector(msg));
2399 crs = crs->next;
2403 void tipc_link_send_duplicate(struct link *l_ptr, struct link *tunnel)
2405 struct sk_buff *iter;
2406 struct tipc_msg tunnel_hdr;
2408 msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2409 DUPLICATE_MSG, TIPC_OK, INT_H_SIZE, l_ptr->addr);
2410 msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2411 msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2412 iter = l_ptr->first_out;
2413 while (iter) {
2414 struct sk_buff *outbuf;
2415 struct tipc_msg *msg = buf_msg(iter);
2416 u32 length = msg_size(msg);
2418 if (msg_user(msg) == MSG_BUNDLER)
2419 msg_set_type(msg, CLOSED_MSG);
2420 msg_set_ack(msg, mod(l_ptr->next_in_no - 1)); /* Update */
2421 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2422 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2423 outbuf = buf_acquire(length + INT_H_SIZE);
2424 if (outbuf == NULL) {
2425 warn("Link changeover error, "
2426 "unable to send duplicate msg\n");
2427 return;
2429 memcpy(outbuf->data, (unchar *)&tunnel_hdr, INT_H_SIZE);
2430 memcpy(outbuf->data + INT_H_SIZE, iter->data, length);
2431 dbg("%c->%c:", l_ptr->b_ptr->net_plane,
2432 tunnel->b_ptr->net_plane);
2433 msg_dbg(buf_msg(outbuf), ">SEND>");
2434 tipc_link_send_buf(tunnel, outbuf);
2435 if (!tipc_link_is_up(l_ptr))
2436 return;
2437 iter = iter->next;
2444 * buf_extract - extracts embedded TIPC message from another message
2445 * @skb: encapsulating message buffer
2446 * @from_pos: offset to extract from
2448 * Returns a new message buffer containing an embedded message. The
2449 * encapsulating message itself is left unchanged.
2452 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2454 struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2455 u32 size = msg_size(msg);
2456 struct sk_buff *eb;
2458 eb = buf_acquire(size);
2459 if (eb)
2460 memcpy(eb->data, (unchar *)msg, size);
2461 return eb;
2465 * link_recv_changeover_msg(): Receive tunneled packet sent
2466 * via other link. Node is locked. Return extracted buffer.
2469 static int link_recv_changeover_msg(struct link **l_ptr,
2470 struct sk_buff **buf)
2472 struct sk_buff *tunnel_buf = *buf;
2473 struct link *dest_link;
2474 struct tipc_msg *msg;
2475 struct tipc_msg *tunnel_msg = buf_msg(tunnel_buf);
2476 u32 msg_typ = msg_type(tunnel_msg);
2477 u32 msg_count = msg_msgcnt(tunnel_msg);
2479 dest_link = (*l_ptr)->owner->links[msg_bearer_id(tunnel_msg)];
2480 if (!dest_link) {
2481 msg_dbg(tunnel_msg, "NOLINK/<REC<");
2482 goto exit;
2484 if (dest_link == *l_ptr) {
2485 err("Unexpected changeover message on link <%s>\n",
2486 (*l_ptr)->name);
2487 goto exit;
2489 dbg("%c<-%c:", dest_link->b_ptr->net_plane,
2490 (*l_ptr)->b_ptr->net_plane);
2491 *l_ptr = dest_link;
2492 msg = msg_get_wrapped(tunnel_msg);
2494 if (msg_typ == DUPLICATE_MSG) {
2495 if (less(msg_seqno(msg), mod(dest_link->next_in_no))) {
2496 msg_dbg(tunnel_msg, "DROP/<REC<");
2497 goto exit;
2499 *buf = buf_extract(tunnel_buf,INT_H_SIZE);
2500 if (*buf == NULL) {
2501 warn("Link changeover error, duplicate msg dropped\n");
2502 goto exit;
2504 msg_dbg(tunnel_msg, "TNL<REC<");
2505 buf_discard(tunnel_buf);
2506 return 1;
2509 /* First original message ?: */
2511 if (tipc_link_is_up(dest_link)) {
2512 msg_dbg(tunnel_msg, "UP/FIRST/<REC<");
2513 info("Resetting link <%s>, changeover initiated by peer\n",
2514 dest_link->name);
2515 tipc_link_reset(dest_link);
2516 dest_link->exp_msg_count = msg_count;
2517 dbg("Expecting %u tunnelled messages\n", msg_count);
2518 if (!msg_count)
2519 goto exit;
2520 } else if (dest_link->exp_msg_count == START_CHANGEOVER) {
2521 msg_dbg(tunnel_msg, "BLK/FIRST/<REC<");
2522 dest_link->exp_msg_count = msg_count;
2523 dbg("Expecting %u tunnelled messages\n", msg_count);
2524 if (!msg_count)
2525 goto exit;
2528 /* Receive original message */
2530 if (dest_link->exp_msg_count == 0) {
2531 warn("Link switchover error, "
2532 "got too many tunnelled messages\n");
2533 msg_dbg(tunnel_msg, "OVERDUE/DROP/<REC<");
2534 dbg_print_link(dest_link, "LINK:");
2535 goto exit;
2537 dest_link->exp_msg_count--;
2538 if (less(msg_seqno(msg), dest_link->reset_checkpoint)) {
2539 msg_dbg(tunnel_msg, "DROP/DUPL/<REC<");
2540 goto exit;
2541 } else {
2542 *buf = buf_extract(tunnel_buf, INT_H_SIZE);
2543 if (*buf != NULL) {
2544 msg_dbg(tunnel_msg, "TNL<REC<");
2545 buf_discard(tunnel_buf);
2546 return 1;
2547 } else {
2548 warn("Link changeover error, original msg dropped\n");
2551 exit:
2552 *buf = NULL;
2553 buf_discard(tunnel_buf);
2554 return 0;
2558 * Bundler functionality:
2560 void tipc_link_recv_bundle(struct sk_buff *buf)
2562 u32 msgcount = msg_msgcnt(buf_msg(buf));
2563 u32 pos = INT_H_SIZE;
2564 struct sk_buff *obuf;
2566 msg_dbg(buf_msg(buf), "<BNDL<: ");
2567 while (msgcount--) {
2568 obuf = buf_extract(buf, pos);
2569 if (obuf == NULL) {
2570 warn("Link unable to unbundle message(s)\n");
2571 break;
2573 pos += align(msg_size(buf_msg(obuf)));
2574 msg_dbg(buf_msg(obuf), " /");
2575 tipc_net_route_msg(obuf);
2577 buf_discard(buf);
2581 * Fragmentation/defragmentation:
2586 * tipc_link_send_long_buf: Entry for buffers needing fragmentation.
2587 * The buffer is complete, inclusive total message length.
2588 * Returns user data length.
2590 int tipc_link_send_long_buf(struct link *l_ptr, struct sk_buff *buf)
2592 struct tipc_msg *inmsg = buf_msg(buf);
2593 struct tipc_msg fragm_hdr;
2594 u32 insize = msg_size(inmsg);
2595 u32 dsz = msg_data_sz(inmsg);
2596 unchar *crs = buf->data;
2597 u32 rest = insize;
2598 u32 pack_sz = link_max_pkt(l_ptr);
2599 u32 fragm_sz = pack_sz - INT_H_SIZE;
2600 u32 fragm_no = 1;
2601 u32 destaddr = msg_destnode(inmsg);
2603 if (msg_short(inmsg))
2604 destaddr = l_ptr->addr;
2606 if (msg_routed(inmsg))
2607 msg_set_prevnode(inmsg, tipc_own_addr);
2609 /* Prepare reusable fragment header: */
2611 msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2612 TIPC_OK, INT_H_SIZE, destaddr);
2613 msg_set_link_selector(&fragm_hdr, msg_link_selector(inmsg));
2614 msg_set_long_msgno(&fragm_hdr, mod(l_ptr->long_msg_seq_no++));
2615 msg_set_fragm_no(&fragm_hdr, fragm_no);
2616 l_ptr->stats.sent_fragmented++;
2618 /* Chop up message: */
2620 while (rest > 0) {
2621 struct sk_buff *fragm;
2623 if (rest <= fragm_sz) {
2624 fragm_sz = rest;
2625 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2627 fragm = buf_acquire(fragm_sz + INT_H_SIZE);
2628 if (fragm == NULL) {
2629 warn("Link unable to fragment message\n");
2630 dsz = -ENOMEM;
2631 goto exit;
2633 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2634 memcpy(fragm->data, (unchar *)&fragm_hdr, INT_H_SIZE);
2635 memcpy(fragm->data + INT_H_SIZE, crs, fragm_sz);
2637 /* Send queued messages first, if any: */
2639 l_ptr->stats.sent_fragments++;
2640 tipc_link_send_buf(l_ptr, fragm);
2641 if (!tipc_link_is_up(l_ptr))
2642 return dsz;
2643 msg_set_fragm_no(&fragm_hdr, ++fragm_no);
2644 rest -= fragm_sz;
2645 crs += fragm_sz;
2646 msg_set_type(&fragm_hdr, FRAGMENT);
2648 exit:
2649 buf_discard(buf);
2650 return dsz;
2654 * A pending message being re-assembled must store certain values
2655 * to handle subsequent fragments correctly. The following functions
2656 * help storing these values in unused, available fields in the
2657 * pending message. This makes dynamic memory allocation unecessary.
2660 static void set_long_msg_seqno(struct sk_buff *buf, u32 seqno)
2662 msg_set_seqno(buf_msg(buf), seqno);
2665 static u32 get_fragm_size(struct sk_buff *buf)
2667 return msg_ack(buf_msg(buf));
2670 static void set_fragm_size(struct sk_buff *buf, u32 sz)
2672 msg_set_ack(buf_msg(buf), sz);
2675 static u32 get_expected_frags(struct sk_buff *buf)
2677 return msg_bcast_ack(buf_msg(buf));
2680 static void set_expected_frags(struct sk_buff *buf, u32 exp)
2682 msg_set_bcast_ack(buf_msg(buf), exp);
2685 static u32 get_timer_cnt(struct sk_buff *buf)
2687 return msg_reroute_cnt(buf_msg(buf));
2690 static void incr_timer_cnt(struct sk_buff *buf)
2692 msg_incr_reroute_cnt(buf_msg(buf));
2696 * tipc_link_recv_fragment(): Called with node lock on. Returns
2697 * the reassembled buffer if message is complete.
2699 int tipc_link_recv_fragment(struct sk_buff **pending, struct sk_buff **fb,
2700 struct tipc_msg **m)
2702 struct sk_buff *prev = NULL;
2703 struct sk_buff *fbuf = *fb;
2704 struct tipc_msg *fragm = buf_msg(fbuf);
2705 struct sk_buff *pbuf = *pending;
2706 u32 long_msg_seq_no = msg_long_msgno(fragm);
2708 *fb = NULL;
2709 msg_dbg(fragm,"FRG<REC<");
2711 /* Is there an incomplete message waiting for this fragment? */
2713 while (pbuf && ((msg_seqno(buf_msg(pbuf)) != long_msg_seq_no)
2714 || (msg_orignode(fragm) != msg_orignode(buf_msg(pbuf))))) {
2715 prev = pbuf;
2716 pbuf = pbuf->next;
2719 if (!pbuf && (msg_type(fragm) == FIRST_FRAGMENT)) {
2720 struct tipc_msg *imsg = (struct tipc_msg *)msg_data(fragm);
2721 u32 msg_sz = msg_size(imsg);
2722 u32 fragm_sz = msg_data_sz(fragm);
2723 u32 exp_fragm_cnt = msg_sz/fragm_sz + !!(msg_sz % fragm_sz);
2724 u32 max = TIPC_MAX_USER_MSG_SIZE + LONG_H_SIZE;
2725 if (msg_type(imsg) == TIPC_MCAST_MSG)
2726 max = TIPC_MAX_USER_MSG_SIZE + MCAST_H_SIZE;
2727 if (msg_size(imsg) > max) {
2728 msg_dbg(fragm,"<REC<Oversized: ");
2729 buf_discard(fbuf);
2730 return 0;
2732 pbuf = buf_acquire(msg_size(imsg));
2733 if (pbuf != NULL) {
2734 pbuf->next = *pending;
2735 *pending = pbuf;
2736 memcpy(pbuf->data, (unchar *)imsg, msg_data_sz(fragm));
2738 /* Prepare buffer for subsequent fragments. */
2740 set_long_msg_seqno(pbuf, long_msg_seq_no);
2741 set_fragm_size(pbuf,fragm_sz);
2742 set_expected_frags(pbuf,exp_fragm_cnt - 1);
2743 } else {
2744 warn("Link unable to reassemble fragmented message\n");
2746 buf_discard(fbuf);
2747 return 0;
2748 } else if (pbuf && (msg_type(fragm) != FIRST_FRAGMENT)) {
2749 u32 dsz = msg_data_sz(fragm);
2750 u32 fsz = get_fragm_size(pbuf);
2751 u32 crs = ((msg_fragm_no(fragm) - 1) * fsz);
2752 u32 exp_frags = get_expected_frags(pbuf) - 1;
2753 memcpy(pbuf->data + crs, msg_data(fragm), dsz);
2754 buf_discard(fbuf);
2756 /* Is message complete? */
2758 if (exp_frags == 0) {
2759 if (prev)
2760 prev->next = pbuf->next;
2761 else
2762 *pending = pbuf->next;
2763 msg_reset_reroute_cnt(buf_msg(pbuf));
2764 *fb = pbuf;
2765 *m = buf_msg(pbuf);
2766 return 1;
2768 set_expected_frags(pbuf,exp_frags);
2769 return 0;
2771 dbg(" Discarding orphan fragment %x\n",fbuf);
2772 msg_dbg(fragm,"ORPHAN:");
2773 dbg("Pending long buffers:\n");
2774 dbg_print_buf_chain(*pending);
2775 buf_discard(fbuf);
2776 return 0;
2780 * link_check_defragm_bufs - flush stale incoming message fragments
2781 * @l_ptr: pointer to link
2784 static void link_check_defragm_bufs(struct link *l_ptr)
2786 struct sk_buff *prev = NULL;
2787 struct sk_buff *next = NULL;
2788 struct sk_buff *buf = l_ptr->defragm_buf;
2790 if (!buf)
2791 return;
2792 if (!link_working_working(l_ptr))
2793 return;
2794 while (buf) {
2795 u32 cnt = get_timer_cnt(buf);
2797 next = buf->next;
2798 if (cnt < 4) {
2799 incr_timer_cnt(buf);
2800 prev = buf;
2801 } else {
2802 dbg(" Discarding incomplete long buffer\n");
2803 msg_dbg(buf_msg(buf), "LONG:");
2804 dbg_print_link(l_ptr, "curr:");
2805 dbg("Pending long buffers:\n");
2806 dbg_print_buf_chain(l_ptr->defragm_buf);
2807 if (prev)
2808 prev->next = buf->next;
2809 else
2810 l_ptr->defragm_buf = buf->next;
2811 buf_discard(buf);
2813 buf = next;
2819 static void link_set_supervision_props(struct link *l_ptr, u32 tolerance)
2821 l_ptr->tolerance = tolerance;
2822 l_ptr->continuity_interval =
2823 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2824 l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2828 void tipc_link_set_queue_limits(struct link *l_ptr, u32 window)
2830 /* Data messages from this node, inclusive FIRST_FRAGM */
2831 l_ptr->queue_limit[DATA_LOW] = window;
2832 l_ptr->queue_limit[DATA_MEDIUM] = (window / 3) * 4;
2833 l_ptr->queue_limit[DATA_HIGH] = (window / 3) * 5;
2834 l_ptr->queue_limit[DATA_CRITICAL] = (window / 3) * 6;
2835 /* Transiting data messages,inclusive FIRST_FRAGM */
2836 l_ptr->queue_limit[DATA_LOW + 4] = 300;
2837 l_ptr->queue_limit[DATA_MEDIUM + 4] = 600;
2838 l_ptr->queue_limit[DATA_HIGH + 4] = 900;
2839 l_ptr->queue_limit[DATA_CRITICAL + 4] = 1200;
2840 l_ptr->queue_limit[CONN_MANAGER] = 1200;
2841 l_ptr->queue_limit[ROUTE_DISTRIBUTOR] = 1200;
2842 l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2843 l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2844 /* FRAGMENT and LAST_FRAGMENT packets */
2845 l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2849 * link_find_link - locate link by name
2850 * @name - ptr to link name string
2851 * @node - ptr to area to be filled with ptr to associated node
2853 * Caller must hold 'tipc_net_lock' to ensure node and bearer are not deleted;
2854 * this also prevents link deletion.
2856 * Returns pointer to link (or 0 if invalid link name).
2859 static struct link *link_find_link(const char *name, struct node **node)
2861 struct link_name link_name_parts;
2862 struct bearer *b_ptr;
2863 struct link *l_ptr;
2865 if (!link_name_validate(name, &link_name_parts))
2866 return NULL;
2868 b_ptr = tipc_bearer_find_interface(link_name_parts.if_local);
2869 if (!b_ptr)
2870 return NULL;
2872 *node = tipc_node_find(link_name_parts.addr_peer);
2873 if (!*node)
2874 return NULL;
2876 l_ptr = (*node)->links[b_ptr->identity];
2877 if (!l_ptr || strcmp(l_ptr->name, name))
2878 return NULL;
2880 return l_ptr;
2883 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2884 u16 cmd)
2886 struct tipc_link_config *args;
2887 u32 new_value;
2888 struct link *l_ptr;
2889 struct node *node;
2890 int res;
2892 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2893 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2895 args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2896 new_value = ntohl(args->value);
2898 if (!strcmp(args->name, tipc_bclink_name)) {
2899 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2900 (tipc_bclink_set_queue_limits(new_value) == 0))
2901 return tipc_cfg_reply_none();
2902 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2903 " (cannot change setting on broadcast link)");
2906 read_lock_bh(&tipc_net_lock);
2907 l_ptr = link_find_link(args->name, &node);
2908 if (!l_ptr) {
2909 read_unlock_bh(&tipc_net_lock);
2910 return tipc_cfg_reply_error_string("link not found");
2913 tipc_node_lock(node);
2914 res = -EINVAL;
2915 switch (cmd) {
2916 case TIPC_CMD_SET_LINK_TOL:
2917 if ((new_value >= TIPC_MIN_LINK_TOL) &&
2918 (new_value <= TIPC_MAX_LINK_TOL)) {
2919 link_set_supervision_props(l_ptr, new_value);
2920 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2921 0, 0, new_value, 0, 0);
2922 res = TIPC_OK;
2924 break;
2925 case TIPC_CMD_SET_LINK_PRI:
2926 if ((new_value >= TIPC_MIN_LINK_PRI) &&
2927 (new_value <= TIPC_MAX_LINK_PRI)) {
2928 l_ptr->priority = new_value;
2929 tipc_link_send_proto_msg(l_ptr, STATE_MSG,
2930 0, 0, 0, new_value, 0);
2931 res = TIPC_OK;
2933 break;
2934 case TIPC_CMD_SET_LINK_WINDOW:
2935 if ((new_value >= TIPC_MIN_LINK_WIN) &&
2936 (new_value <= TIPC_MAX_LINK_WIN)) {
2937 tipc_link_set_queue_limits(l_ptr, new_value);
2938 res = TIPC_OK;
2940 break;
2942 tipc_node_unlock(node);
2944 read_unlock_bh(&tipc_net_lock);
2945 if (res)
2946 return tipc_cfg_reply_error_string("cannot change link setting");
2948 return tipc_cfg_reply_none();
2952 * link_reset_statistics - reset link statistics
2953 * @l_ptr: pointer to link
2956 static void link_reset_statistics(struct link *l_ptr)
2958 memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2959 l_ptr->stats.sent_info = l_ptr->next_out_no;
2960 l_ptr->stats.recv_info = l_ptr->next_in_no;
2963 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2965 char *link_name;
2966 struct link *l_ptr;
2967 struct node *node;
2969 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2970 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2972 link_name = (char *)TLV_DATA(req_tlv_area);
2973 if (!strcmp(link_name, tipc_bclink_name)) {
2974 if (tipc_bclink_reset_stats())
2975 return tipc_cfg_reply_error_string("link not found");
2976 return tipc_cfg_reply_none();
2979 read_lock_bh(&tipc_net_lock);
2980 l_ptr = link_find_link(link_name, &node);
2981 if (!l_ptr) {
2982 read_unlock_bh(&tipc_net_lock);
2983 return tipc_cfg_reply_error_string("link not found");
2986 tipc_node_lock(node);
2987 link_reset_statistics(l_ptr);
2988 tipc_node_unlock(node);
2989 read_unlock_bh(&tipc_net_lock);
2990 return tipc_cfg_reply_none();
2994 * percent - convert count to a percentage of total (rounding up or down)
2997 static u32 percent(u32 count, u32 total)
2999 return (count * 100 + (total / 2)) / total;
3003 * tipc_link_stats - print link statistics
3004 * @name: link name
3005 * @buf: print buffer area
3006 * @buf_size: size of print buffer area
3008 * Returns length of print buffer data string (or 0 if error)
3011 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
3013 struct print_buf pb;
3014 struct link *l_ptr;
3015 struct node *node;
3016 char *status;
3017 u32 profile_total = 0;
3019 if (!strcmp(name, tipc_bclink_name))
3020 return tipc_bclink_stats(buf, buf_size);
3022 tipc_printbuf_init(&pb, buf, buf_size);
3024 read_lock_bh(&tipc_net_lock);
3025 l_ptr = link_find_link(name, &node);
3026 if (!l_ptr) {
3027 read_unlock_bh(&tipc_net_lock);
3028 return 0;
3030 tipc_node_lock(node);
3032 if (tipc_link_is_active(l_ptr))
3033 status = "ACTIVE";
3034 else if (tipc_link_is_up(l_ptr))
3035 status = "STANDBY";
3036 else
3037 status = "DEFUNCT";
3038 tipc_printf(&pb, "Link <%s>\n"
3039 " %s MTU:%u Priority:%u Tolerance:%u ms"
3040 " Window:%u packets\n",
3041 l_ptr->name, status, link_max_pkt(l_ptr),
3042 l_ptr->priority, l_ptr->tolerance, l_ptr->queue_limit[0]);
3043 tipc_printf(&pb, " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
3044 l_ptr->next_in_no - l_ptr->stats.recv_info,
3045 l_ptr->stats.recv_fragments,
3046 l_ptr->stats.recv_fragmented,
3047 l_ptr->stats.recv_bundles,
3048 l_ptr->stats.recv_bundled);
3049 tipc_printf(&pb, " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
3050 l_ptr->next_out_no - l_ptr->stats.sent_info,
3051 l_ptr->stats.sent_fragments,
3052 l_ptr->stats.sent_fragmented,
3053 l_ptr->stats.sent_bundles,
3054 l_ptr->stats.sent_bundled);
3055 profile_total = l_ptr->stats.msg_length_counts;
3056 if (!profile_total)
3057 profile_total = 1;
3058 tipc_printf(&pb, " TX profile sample:%u packets average:%u octets\n"
3059 " 0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
3060 "-16354:%u%% -32768:%u%% -66000:%u%%\n",
3061 l_ptr->stats.msg_length_counts,
3062 l_ptr->stats.msg_lengths_total / profile_total,
3063 percent(l_ptr->stats.msg_length_profile[0], profile_total),
3064 percent(l_ptr->stats.msg_length_profile[1], profile_total),
3065 percent(l_ptr->stats.msg_length_profile[2], profile_total),
3066 percent(l_ptr->stats.msg_length_profile[3], profile_total),
3067 percent(l_ptr->stats.msg_length_profile[4], profile_total),
3068 percent(l_ptr->stats.msg_length_profile[5], profile_total),
3069 percent(l_ptr->stats.msg_length_profile[6], profile_total));
3070 tipc_printf(&pb, " RX states:%u probes:%u naks:%u defs:%u dups:%u\n",
3071 l_ptr->stats.recv_states,
3072 l_ptr->stats.recv_probes,
3073 l_ptr->stats.recv_nacks,
3074 l_ptr->stats.deferred_recv,
3075 l_ptr->stats.duplicates);
3076 tipc_printf(&pb, " TX states:%u probes:%u naks:%u acks:%u dups:%u\n",
3077 l_ptr->stats.sent_states,
3078 l_ptr->stats.sent_probes,
3079 l_ptr->stats.sent_nacks,
3080 l_ptr->stats.sent_acks,
3081 l_ptr->stats.retransmitted);
3082 tipc_printf(&pb, " Congestion bearer:%u link:%u Send queue max:%u avg:%u\n",
3083 l_ptr->stats.bearer_congs,
3084 l_ptr->stats.link_congs,
3085 l_ptr->stats.max_queue_sz,
3086 l_ptr->stats.queue_sz_counts
3087 ? (l_ptr->stats.accu_queue_sz / l_ptr->stats.queue_sz_counts)
3088 : 0);
3090 tipc_node_unlock(node);
3091 read_unlock_bh(&tipc_net_lock);
3092 return tipc_printbuf_validate(&pb);
3095 #define MAX_LINK_STATS_INFO 2000
3097 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
3099 struct sk_buff *buf;
3100 struct tlv_desc *rep_tlv;
3101 int str_len;
3103 if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
3104 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
3106 buf = tipc_cfg_reply_alloc(TLV_SPACE(MAX_LINK_STATS_INFO));
3107 if (!buf)
3108 return NULL;
3110 rep_tlv = (struct tlv_desc *)buf->data;
3112 str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
3113 (char *)TLV_DATA(rep_tlv), MAX_LINK_STATS_INFO);
3114 if (!str_len) {
3115 buf_discard(buf);
3116 return tipc_cfg_reply_error_string("link not found");
3119 skb_put(buf, TLV_SPACE(str_len));
3120 TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
3122 return buf;
3125 #if 0
3126 int link_control(const char *name, u32 op, u32 val)
3128 int res = -EINVAL;
3129 struct link *l_ptr;
3130 u32 bearer_id;
3131 struct node * node;
3132 u32 a;
3134 a = link_name2addr(name, &bearer_id);
3135 read_lock_bh(&tipc_net_lock);
3136 node = tipc_node_find(a);
3137 if (node) {
3138 tipc_node_lock(node);
3139 l_ptr = node->links[bearer_id];
3140 if (l_ptr) {
3141 if (op == TIPC_REMOVE_LINK) {
3142 struct bearer *b_ptr = l_ptr->b_ptr;
3143 spin_lock_bh(&b_ptr->publ.lock);
3144 tipc_link_delete(l_ptr);
3145 spin_unlock_bh(&b_ptr->publ.lock);
3147 if (op == TIPC_CMD_BLOCK_LINK) {
3148 tipc_link_reset(l_ptr);
3149 l_ptr->blocked = 1;
3151 if (op == TIPC_CMD_UNBLOCK_LINK) {
3152 l_ptr->blocked = 0;
3154 res = TIPC_OK;
3156 tipc_node_unlock(node);
3158 read_unlock_bh(&tipc_net_lock);
3159 return res;
3161 #endif
3164 * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
3165 * @dest: network address of destination node
3166 * @selector: used to select from set of active links
3168 * If no active link can be found, uses default maximum packet size.
3171 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
3173 struct node *n_ptr;
3174 struct link *l_ptr;
3175 u32 res = MAX_PKT_DEFAULT;
3177 if (dest == tipc_own_addr)
3178 return MAX_MSG_SIZE;
3180 read_lock_bh(&tipc_net_lock);
3181 n_ptr = tipc_node_select(dest, selector);
3182 if (n_ptr) {
3183 tipc_node_lock(n_ptr);
3184 l_ptr = n_ptr->active_links[selector & 1];
3185 if (l_ptr)
3186 res = link_max_pkt(l_ptr);
3187 tipc_node_unlock(n_ptr);
3189 read_unlock_bh(&tipc_net_lock);
3190 return res;
3193 #if 0
3194 static void link_dump_rec_queue(struct link *l_ptr)
3196 struct sk_buff *crs;
3198 if (!l_ptr->oldest_deferred_in) {
3199 info("Reception queue empty\n");
3200 return;
3202 info("Contents of Reception queue:\n");
3203 crs = l_ptr->oldest_deferred_in;
3204 while (crs) {
3205 if (crs->data == (void *)0x0000a3a3) {
3206 info("buffer %x invalid\n", crs);
3207 return;
3209 msg_dbg(buf_msg(crs), "In rec queue: \n");
3210 crs = crs->next;
3213 #endif
3215 static void link_dump_send_queue(struct link *l_ptr)
3217 if (l_ptr->next_out) {
3218 info("\nContents of unsent queue:\n");
3219 dbg_print_buf_chain(l_ptr->next_out);
3221 info("\nContents of send queue:\n");
3222 if (l_ptr->first_out) {
3223 dbg_print_buf_chain(l_ptr->first_out);
3225 info("Empty send queue\n");
3228 static void link_print(struct link *l_ptr, struct print_buf *buf,
3229 const char *str)
3231 tipc_printf(buf, str);
3232 if (link_reset_reset(l_ptr) || link_reset_unknown(l_ptr))
3233 return;
3234 tipc_printf(buf, "Link %x<%s>:",
3235 l_ptr->addr, l_ptr->b_ptr->publ.name);
3236 tipc_printf(buf, ": NXO(%u):", mod(l_ptr->next_out_no));
3237 tipc_printf(buf, "NXI(%u):", mod(l_ptr->next_in_no));
3238 tipc_printf(buf, "SQUE");
3239 if (l_ptr->first_out) {
3240 tipc_printf(buf, "[%u..", msg_seqno(buf_msg(l_ptr->first_out)));
3241 if (l_ptr->next_out)
3242 tipc_printf(buf, "%u..",
3243 msg_seqno(buf_msg(l_ptr->next_out)));
3244 tipc_printf(buf, "%u]",
3245 msg_seqno(buf_msg
3246 (l_ptr->last_out)), l_ptr->out_queue_size);
3247 if ((mod(msg_seqno(buf_msg(l_ptr->last_out)) -
3248 msg_seqno(buf_msg(l_ptr->first_out)))
3249 != (l_ptr->out_queue_size - 1))
3250 || (l_ptr->last_out->next != 0)) {
3251 tipc_printf(buf, "\nSend queue inconsistency\n");
3252 tipc_printf(buf, "first_out= %x ", l_ptr->first_out);
3253 tipc_printf(buf, "next_out= %x ", l_ptr->next_out);
3254 tipc_printf(buf, "last_out= %x ", l_ptr->last_out);
3255 link_dump_send_queue(l_ptr);
3257 } else
3258 tipc_printf(buf, "[]");
3259 tipc_printf(buf, "SQSIZ(%u)", l_ptr->out_queue_size);
3260 if (l_ptr->oldest_deferred_in) {
3261 u32 o = msg_seqno(buf_msg(l_ptr->oldest_deferred_in));
3262 u32 n = msg_seqno(buf_msg(l_ptr->newest_deferred_in));
3263 tipc_printf(buf, ":RQUE[%u..%u]", o, n);
3264 if (l_ptr->deferred_inqueue_sz != mod((n + 1) - o)) {
3265 tipc_printf(buf, ":RQSIZ(%u)",
3266 l_ptr->deferred_inqueue_sz);
3269 if (link_working_unknown(l_ptr))
3270 tipc_printf(buf, ":WU");
3271 if (link_reset_reset(l_ptr))
3272 tipc_printf(buf, ":RR");
3273 if (link_reset_unknown(l_ptr))
3274 tipc_printf(buf, ":RU");
3275 if (link_working_working(l_ptr))
3276 tipc_printf(buf, ":WW");
3277 tipc_printf(buf, "\n");