2 /* Copyright (c) 2010-2013. The SimGrid Team.
3 * All rights reserved. */
5 /* This program is free software; you can redistribute it and/or modify it
6 * under the terms of the license (GNU LGPL) which comes with this package. */
11 #include "xbt/asserts.h"
12 #include "simgrid/modelchecker.h"
13 #include <xbt/RngStream.h>
15 /** @addtogroup MSG_examples
17 * - <b>chord/chord.c: Classical Chord P2P protocol</b>
18 * This example implements the well known Chord P2P protocol. Its
19 * main advantage is that it constitute a fully working non-trivial
20 * example. In addition, its implementation is rather efficient, as
21 * demonstrated in http://hal.inria.fr/inria-00602216/
25 XBT_LOG_NEW_DEFAULT_CATEGORY(msg_chord,
26 "Messages specific for this msg example");
30 #define MAILBOX_NAME_SIZE 10
32 static int nb_bits = 24;
33 static int nb_keys = 0;
34 static int timeout = 50;
35 static int max_simulation_time = 1000;
36 static int periodic_stabilize_delay = 20;
37 static int periodic_fix_fingers_delay = 120;
38 static int periodic_check_predecessor_delay = 120;
39 static int periodic_lookup_delay = 10;
41 extern long int smx_total_comms;
46 typedef struct s_finger {
48 char mailbox[MAILBOX_NAME_SIZE]; // string representation of the id
49 } s_finger_t, *finger_t;
54 typedef struct s_node {
56 char mailbox[MAILBOX_NAME_SIZE]; // my mailbox name (string representation of the id)
57 s_finger_t *fingers; // finger table, of size nb_bits (fingers[0] is my successor)
58 int pred_id; // predecessor id
59 char pred_mailbox[MAILBOX_NAME_SIZE]; // predecessor's mailbox name
60 int next_finger_to_fix; // index of the next finger to fix in fix_fingers()
61 msg_comm_t comm_receive; // current communication to receive
62 double last_change_date; // last time I changed a finger or my predecessor
63 RngStream stream; //RngStream for
67 * Types of tasks exchanged between nodes.
71 TASK_FIND_SUCCESSOR_ANSWER,
73 TASK_GET_PREDECESSOR_ANSWER,
75 TASK_SUCCESSOR_LEAVING,
76 TASK_PREDECESSOR_LEAVING,
77 TASK_PREDECESSOR_ALIVE,
78 TASK_PREDECESSOR_ALIVE_ANSWER
82 * Data attached with the tasks sent and received
84 typedef struct s_task_data {
85 e_task_type_t type; // type of task
86 int request_id; // id paramater (used by some types of tasks)
87 int request_finger; // finger parameter (used by some types of tasks)
88 int answer_id; // answer (used by some types of tasks)
89 char answer_to[MAILBOX_NAME_SIZE]; // mailbox to send an answer to (if any)
90 const char* issuer_host_name; // used for logging
91 } s_task_data_t, *task_data_t;
96 static void chord_initialize(void);
97 static void chord_exit(void);
98 static int normalize(int id);
99 static int is_in_interval(int id, int start, int end);
100 static void get_mailbox(int host_id, char* mailbox);
101 static void task_free(void* task);
102 static void print_finger_table(node_t node);
103 static void set_finger(node_t node, int finger_index, int id);
104 static void set_predecessor(node_t node, int predecessor_id);
107 static int node(int argc, char *argv[]);
108 static void handle_task(node_t node, msg_task_t task);
111 static void create(node_t node);
112 static int join(node_t node, int known_id);
113 static void leave(node_t node);
114 static int find_successor(node_t node, int id);
115 static int remote_find_successor(node_t node, int ask_to_id, int id);
116 static int remote_get_predecessor(node_t node, int ask_to_id);
117 static int closest_preceding_node(node_t node, int id);
118 static void stabilize(node_t node);
119 static void notify(node_t node, int predecessor_candidate_id);
120 static void remote_notify(node_t node, int notify_to, int predecessor_candidate_id);
121 static void fix_fingers(node_t node);
122 static void check_predecessor(node_t node);
123 static void random_lookup(node_t);
124 static void quit_notify(node_t node);
127 * \brief Global initialization of the Chord simulation.
129 static void chord_initialize(void)
131 // compute the powers of 2 once for all
132 powers2 = xbt_new(int, nb_bits);
135 for (i = 0; i < nb_bits; i++) {
140 XBT_DEBUG("Sets nb_keys to %d", nb_keys);
143 static void chord_exit(void)
149 * \brief Turns an id into an equivalent id in [0, nb_keys).
151 * \return the corresponding normalized id
153 static int normalize(int id)
155 // like id % nb_keys, but works with negatives numbers (and faster)
156 return id & (nb_keys - 1);
160 * \brief Returns whether an id belongs to the interval [start, end].
162 * The parameters are noramlized to make sure they are between 0 and nb_keys - 1).
163 * 1 belongs to [62, 3]
164 * 1 does not belong to [3, 62]
165 * 63 belongs to [62, 3]
166 * 63 does not belong to [3, 62]
167 * 24 belongs to [21, 29]
168 * 24 does not belong to [29, 21]
170 * \param id id to check
171 * \param start lower bound
172 * \param end upper bound
173 * \return a non-zero value if id in in [start, end]
175 static int is_in_interval(int id, int start, int end)
178 start = normalize(start);
179 end = normalize(end);
181 // make sure end >= start and id >= start
194 * \brief Gets the mailbox name of a host given its chord id.
195 * \param node_id id of a node
196 * \param mailbox pointer to where the mailbox name should be written
197 * (there must be enough space)
199 static void get_mailbox(int node_id, char* mailbox)
201 snprintf(mailbox, MAILBOX_NAME_SIZE - 1, "%d", node_id);
205 * \brief Frees the memory used by a task.
206 * \param task the MSG task to destroy
208 static void task_free(void* task)
210 // TODO add a parameter data_free_function to MSG_task_create?
212 xbt_free(MSG_task_get_data(task));
213 MSG_task_destroy(task);
218 * \brief Displays the finger table of a node.
221 static void print_finger_table(node_t node)
223 if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) {
225 XBT_VERB("My finger table:");
226 XBT_VERB("Start | Succ ");
227 for (i = 0; i < nb_bits; i++) {
228 XBT_VERB(" %3d | %3d ", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
230 XBT_VERB("Predecessor: %d", node->pred_id);
235 * \brief Sets a finger of the current node.
236 * \param node the current node
237 * \param finger_index index of the finger to set (0 to nb_bits - 1)
238 * \param id the id to set for this finger
240 static void set_finger(node_t node, int finger_index, int id)
242 if (id != node->fingers[finger_index].id) {
243 node->fingers[finger_index].id = id;
244 get_mailbox(id, node->fingers[finger_index].mailbox);
245 node->last_change_date = MSG_get_clock();
246 XBT_DEBUG("My new finger #%d is %d", finger_index, id);
251 * \brief Sets the predecessor of the current node.
252 * \param node the current node
253 * \param id the id to predecessor, or -1 to unset the predecessor
255 static void set_predecessor(node_t node, int predecessor_id)
257 if (predecessor_id != node->pred_id) {
258 node->pred_id = predecessor_id;
260 if (predecessor_id != -1) {
261 get_mailbox(predecessor_id, node->pred_mailbox);
263 node->last_change_date = MSG_get_clock();
265 XBT_DEBUG("My new predecessor is %d", predecessor_id);
270 * \brief Node Function
273 * - the id of a guy I know in the system (except for the first node)
274 * - the time to sleep before I join (except for the first node)
276 int node(int argc, char *argv[])
279 /* Reduce the run size for the MC */
281 periodic_stabilize_delay = 8;
282 periodic_fix_fingers_delay = 8;
283 periodic_check_predecessor_delay = 8;
286 double init_time = MSG_get_clock();
287 msg_task_t task_received = NULL;
289 int join_success = 0;
291 double next_stabilize_date = init_time + periodic_stabilize_delay;
292 double next_fix_fingers_date = init_time + periodic_fix_fingers_delay;
293 double next_check_predecessor_date = init_time + periodic_check_predecessor_delay;
294 double next_lookup_date = init_time + periodic_lookup_delay;
298 int sub_protocol = 0;
300 xbt_assert(argc == 3 || argc == 5, "Wrong number of arguments for this node");
302 // initialize my node
304 node.id = atoi(argv[1]);
305 get_mailbox(node.id, node.mailbox);
306 node.next_finger_to_fix = 0;
307 node.fingers = xbt_new0(s_finger_t, nb_bits);
308 node.last_change_date = init_time;
310 for (i = 0; i < nb_bits; i++) {
311 node.fingers[i].id = -1;
312 set_finger(&node, i, node.id);
315 if (argc == 3) { // first ring
316 deadline = atof(argv[2]);
321 int known_id = atoi(argv[2]);
322 //double sleep_time = atof(argv[3]);
323 deadline = atof(argv[4]);
326 // sleep before starting
327 XBT_DEBUG("Let's sleep during %f", sleep_time);
328 MSG_process_sleep(sleep_time);
330 XBT_DEBUG("Hey! Let's join the system.");
332 join_success = join(&node, known_id);
336 while (MSG_get_clock() < init_time + deadline
337 // && MSG_get_clock() < node.last_change_date + 1000
338 && MSG_get_clock() < max_simulation_time) {
340 if (node.comm_receive == NULL) {
341 task_received = NULL;
342 node.comm_receive = MSG_task_irecv(&task_received, node.mailbox);
343 // FIXME: do not make MSG_task_irecv() calls from several functions
346 //XBT_INFO("Node %d is ring member : %d", node.id, is_ring_member(known_id, node.id) != -1);
348 if (!MSG_comm_test(node.comm_receive)) {
350 // no task was received: make some periodic calls
353 if(!MC_visited_reduction() && no_op){
356 if(listen == 0 && (sub_protocol = MC_random(0, 4)) > 0){
357 if(sub_protocol == 1)
359 else if(sub_protocol == 2)
361 else if(sub_protocol == 3)
362 check_predecessor(&node);
364 random_lookup(&node);
367 MSG_process_sleep(5);
368 if(!MC_visited_reduction())
372 if (MSG_get_clock() >= next_stabilize_date) {
374 next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay;
375 }else if (MSG_get_clock() >= next_fix_fingers_date) {
377 next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay;
378 }else if (MSG_get_clock() >= next_check_predecessor_date) {
379 check_predecessor(&node);
380 next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay;
381 }else if (MSG_get_clock() >= next_lookup_date) {
382 random_lookup(&node);
383 next_lookup_date = MSG_get_clock() + periodic_lookup_delay;
385 // nothing to do: sleep for a while
386 MSG_process_sleep(5);
391 // a transfer has occurred
393 msg_error_t status = MSG_comm_get_status(node.comm_receive);
395 if (status != MSG_OK) {
396 XBT_DEBUG("Failed to receive a task. Nevermind.");
397 MSG_comm_destroy(node.comm_receive);
398 node.comm_receive = NULL;
401 // the task was successfully received
402 MSG_comm_destroy(node.comm_receive);
403 node.comm_receive = NULL;
404 handle_task(&node, task_received);
409 if (node.comm_receive) {
410 MSG_comm_destroy(node.comm_receive);
411 node.comm_receive = NULL;
418 // stop the simulation
419 xbt_free(node.fingers);
424 * \brief This function is called when the current node receives a task.
425 * \param node the current node
426 * \param task the task to handle (don't touch it then:
427 * it will be destroyed, reused or forwarded)
429 static void handle_task(node_t node, msg_task_t task) {
431 XBT_DEBUG("Handling task %p", task);
432 char mailbox[MAILBOX_NAME_SIZE];
433 task_data_t task_data = (task_data_t) MSG_task_get_data(task);
434 e_task_type_t type = task_data->type;
438 case TASK_FIND_SUCCESSOR:
439 XBT_DEBUG("Receiving a 'Find Successor' request from %s for id %d",
440 task_data->issuer_host_name, task_data->request_id);
441 // is my successor the successor?
442 if (is_in_interval(task_data->request_id, node->id + 1, node->fingers[0].id)) {
443 task_data->type = TASK_FIND_SUCCESSOR_ANSWER;
444 task_data->answer_id = node->fingers[0].id;
445 XBT_DEBUG("Sending back a 'Find Successor Answer' to %s (mailbox %s): the successor of %d is %d",
446 task_data->issuer_host_name,
447 task_data->answer_to,
448 task_data->request_id, task_data->answer_id);
449 MSG_task_dsend(task, task_data->answer_to, task_free);
452 // otherwise, forward the request to the closest preceding finger in my table
453 int closest = closest_preceding_node(node, task_data->request_id);
454 XBT_DEBUG("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d",
455 task_data->request_id, closest);
456 get_mailbox(closest, mailbox);
457 MSG_task_dsend(task, mailbox, task_free);
461 case TASK_GET_PREDECESSOR:
462 XBT_DEBUG("Receiving a 'Get Predecessor' request from %s", task_data->issuer_host_name);
463 task_data->type = TASK_GET_PREDECESSOR_ANSWER;
464 task_data->answer_id = node->pred_id;
465 XBT_DEBUG("Sending back a 'Get Predecessor Answer' to %s via mailbox '%s': my predecessor is %d",
466 task_data->issuer_host_name,
467 task_data->answer_to, task_data->answer_id);
468 MSG_task_dsend(task, task_data->answer_to, task_free);
472 // someone is telling me that he may be my new predecessor
473 XBT_DEBUG("Receiving a 'Notify' request from %s", task_data->issuer_host_name);
474 notify(node, task_data->request_id);
478 case TASK_PREDECESSOR_LEAVING:
479 // my predecessor is about to quit
480 XBT_DEBUG("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name);
481 // modify my predecessor
482 set_predecessor(node, task_data->request_id);
485 >> notify my new predecessor
486 >> send a notify_predecessors !!
490 case TASK_SUCCESSOR_LEAVING:
491 // my successor is about to quit
492 XBT_DEBUG("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name);
493 // modify my successor FIXME : this should be implicit ?
494 set_finger(node, 0, task_data->request_id);
497 >> notify my new successor
498 >> update my table & predecessors table */
501 case TASK_FIND_SUCCESSOR_ANSWER:
502 case TASK_GET_PREDECESSOR_ANSWER:
503 case TASK_PREDECESSOR_ALIVE_ANSWER:
504 XBT_DEBUG("Ignoring unexpected task of type %d (%p)", (int)type, task);
508 case TASK_PREDECESSOR_ALIVE:
509 XBT_DEBUG("Receiving a 'Predecessor Alive' request from %s", task_data->issuer_host_name);
510 task_data->type = TASK_PREDECESSOR_ALIVE_ANSWER;
511 XBT_DEBUG("Sending back a 'Predecessor Alive Answer' to %s (mailbox %s)",
512 task_data->issuer_host_name,
513 task_data->answer_to);
514 MSG_task_dsend(task, task_data->answer_to, task_free);
521 * \brief Initializes the current node as the first one of the system.
522 * \param node the current node
524 static void create(node_t node)
526 XBT_DEBUG("Create a new Chord ring...");
527 set_predecessor(node, -1); // -1 means that I have no predecessor
528 print_finger_table(node);
532 * \brief Makes the current node join the ring, knowing the id of a node
533 * already in the ring
534 * \param node the current node
535 * \param known_id id of a node already in the ring
536 * \return 1 if the join operation succeeded, 0 otherwise
538 static int join(node_t node, int known_id)
540 XBT_INFO("Joining the ring with id %d, knowing node %d", node->id, known_id);
541 set_predecessor(node, -1); // no predecessor (yet)
545 for (i = 0; i < nb_bits; i++) {
546 set_finger(node, i, known_id);
550 int successor_id = remote_find_successor(node, known_id, node->id);
551 if (successor_id == -1) {
552 XBT_INFO("Cannot join the ring.");
555 set_finger(node, 0, successor_id);
556 print_finger_table(node);
559 return successor_id != -1;
563 * \brief Makes the current node quit the system
564 * \param node the current node
566 static void leave(node_t node)
568 XBT_DEBUG("Well Guys! I Think it's time for me to quit ;)");
570 RngStream_DeleteStream(&node->stream);
574 * \brief Notifies the successor and the predecessor of the current node
576 * \param node the current node
578 static void quit_notify(node_t node)
580 char mailbox[MAILBOX_NAME_SIZE];
581 //send the PREDECESSOR_LEAVING to our successor
582 task_data_t req_data = xbt_new0(s_task_data_t,1);
583 req_data->type = TASK_PREDECESSOR_LEAVING;
584 req_data->request_id = node->pred_id;
585 get_mailbox(node->id, req_data->answer_to);
586 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
588 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
589 XBT_DEBUG("Sending a 'PREDECESSOR_LEAVING' to my successor %d",node->fingers[0].id);
590 if (MSG_task_send_with_timeout(task_sent, node->fingers[0].mailbox, timeout)==
592 XBT_DEBUG("Timeout expired when sending a 'PREDECESSOR_LEAVING' to my successor %d",
593 node->fingers[0].id);
594 task_free(task_sent);
597 //send the SUCCESSOR_LEAVING to our predecessor
598 get_mailbox(node->pred_id, mailbox);
599 task_data_t req_data_s = xbt_new0(s_task_data_t,1);
600 req_data_s->type = TASK_SUCCESSOR_LEAVING;
601 req_data_s->request_id = node->fingers[0].id;
602 req_data_s->request_id = node->pred_id;
603 get_mailbox(node->id, req_data_s->answer_to);
604 req_data_s->issuer_host_name = MSG_host_get_name(MSG_host_self());
606 msg_task_t task_sent_s = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data_s);
607 XBT_DEBUG("Sending a 'SUCCESSOR_LEAVING' to my predecessor %d",node->pred_id);
608 if (MSG_task_send_with_timeout(task_sent_s, mailbox, timeout)==
610 XBT_DEBUG("Timeout expired when sending a 'SUCCESSOR_LEAVING' to my predecessor %d",
612 task_free(task_sent_s);
618 * \brief Makes the current node find the successor node of an id.
619 * \param node the current node
620 * \param id the id to find
621 * \return the id of the successor node, or -1 if the request failed
623 static int find_successor(node_t node, int id)
625 // is my successor the successor?
626 if (is_in_interval(id, node->id + 1, node->fingers[0].id)) {
627 return node->fingers[0].id;
630 // otherwise, ask the closest preceding finger in my table
631 int closest = closest_preceding_node(node, id);
632 return remote_find_successor(node, closest, id);
636 * \brief Asks another node the successor node of an id.
637 * \param node the current node
638 * \param ask_to the node to ask to
639 * \param id the id to find
640 * \return the id of the successor node, or -1 if the request failed
642 static int remote_find_successor(node_t node, int ask_to, int id)
646 char mailbox[MAILBOX_NAME_SIZE];
647 get_mailbox(ask_to, mailbox);
648 task_data_t req_data = xbt_new0(s_task_data_t, 1);
649 req_data->type = TASK_FIND_SUCCESSOR;
650 req_data->request_id = id;
651 get_mailbox(node->id, req_data->answer_to);
652 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
654 // send a "Find Successor" request to ask_to_id
655 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
656 XBT_DEBUG("Sending a 'Find Successor' request (task %p) to %d for id %d", task_sent, ask_to, id);
657 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
660 XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
661 task_sent, ask_to, id);
662 task_free(task_sent);
666 // receive the answer
667 XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
668 task_sent, ask_to, id);
671 if (node->comm_receive == NULL) {
672 msg_task_t task_received = NULL;
673 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
676 res = MSG_comm_wait(node->comm_receive, timeout);
679 XBT_DEBUG("Failed to receive the answer to my 'Find Successor' request (task %p): %d",
680 task_sent, (int)res);
682 MSG_comm_destroy(node->comm_receive);
683 node->comm_receive = NULL;
686 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
687 XBT_DEBUG("Received a task (%p)", task_received);
688 task_data_t ans_data = MSG_task_get_data(task_received);
690 // Once upon a time, our code assumed that here, task_received != task_sent all the time
692 // This assumption is wrong (as messages from differing round can interleave), leading to a bug in our code.
693 // We failed to find this bug directly, as it only occured on large platforms, leading to hardly usable traces.
694 // Instead, we used the model-checker to track down the issue by adding the following test here in the code:
695 // if (MC_is_active()) {
696 // MC_assert(task_received == task_sent);
698 // That explained the bug in a snap, with a very cool example and everything.
700 // This MC_assert is now desactivated as the case is now properly handled in our code and we don't want the
701 // MC to fail any further under that condition, but this comment is here to as a memorial for this first
702 // brillant victory of the model-checking in the SimGrid community :)
704 if (task_received != task_sent) {
705 // this is not the expected answer
706 MSG_comm_destroy(node->comm_receive);
707 node->comm_receive = NULL;
708 handle_task(node, task_received);
711 // this is our answer
712 XBT_DEBUG("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d",
713 ans_data->request_id, task_received, id, ans_data->answer_id);
714 successor = ans_data->answer_id;
716 MSG_comm_destroy(node->comm_receive);
717 node->comm_receive = NULL;
718 task_free(task_received);
728 * \brief Asks another node its predecessor.
729 * \param node the current node
730 * \param ask_to the node to ask to
731 * \return the id of its predecessor node, or -1 if the request failed
732 * (or if the node does not know its predecessor)
734 static int remote_get_predecessor(node_t node, int ask_to)
736 int predecessor_id = -1;
738 char mailbox[MAILBOX_NAME_SIZE];
739 get_mailbox(ask_to, mailbox);
740 task_data_t req_data = xbt_new0(s_task_data_t, 1);
741 req_data->type = TASK_GET_PREDECESSOR;
742 get_mailbox(node->id, req_data->answer_to);
743 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
745 // send a "Get Predecessor" request to ask_to_id
746 XBT_DEBUG("Sending a 'Get Predecessor' request to %d", ask_to);
747 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
748 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
751 XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d",
753 task_free(task_sent);
757 // receive the answer
758 XBT_DEBUG("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
759 task_sent, ask_to, req_data->answer_to);
762 if (node->comm_receive == NULL) { // FIXME simplify this
763 msg_task_t task_received = NULL;
764 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
767 res = MSG_comm_wait(node->comm_receive, timeout);
770 XBT_DEBUG("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d",
771 task_sent, (int)res);
773 MSG_comm_destroy(node->comm_receive);
774 node->comm_receive = NULL;
775 task_free(task_sent);
778 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
779 task_data_t ans_data = MSG_task_get_data(task_received);
781 /*if (MC_is_active()) {
782 MC_assert(task_received == task_sent);
785 if (task_received != task_sent) {
786 MSG_comm_destroy(node->comm_receive);
787 node->comm_receive = NULL;
788 handle_task(node, task_received);
791 XBT_DEBUG("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d",
792 task_received, ask_to, ans_data->answer_id);
793 predecessor_id = ans_data->answer_id;
795 MSG_comm_destroy(node->comm_receive);
796 node->comm_receive = NULL;
797 task_free(task_received);
803 return predecessor_id;
807 * \brief Returns the closest preceding finger of an id
808 * with respect to the finger table of the current node.
809 * \param node the current node
810 * \param id the id to find
811 * \return the closest preceding finger of that id
813 int closest_preceding_node(node_t node, int id)
816 for (i = nb_bits - 1; i >= 0; i--) {
817 if (is_in_interval(node->fingers[i].id, node->id + 1, id - 1)) {
818 return node->fingers[i].id;
825 * \brief This function is called periodically. It checks the immediate
826 * successor of the current node.
827 * \param node the current node
829 static void stabilize(node_t node)
831 XBT_DEBUG("Stabilizing node");
833 // get the predecessor of my immediate successor
835 int successor_id = node->fingers[0].id;
836 if (successor_id != node->id) {
837 candidate_id = remote_get_predecessor(node, successor_id);
840 candidate_id = node->pred_id;
843 // this node is a candidate to become my new successor
844 if (candidate_id != -1
845 && is_in_interval(candidate_id, node->id + 1, successor_id - 1)) {
846 set_finger(node, 0, candidate_id);
848 if (successor_id != node->id) {
849 remote_notify(node, successor_id, node->id);
854 * \brief Notifies the current node that its predecessor may have changed.
855 * \param node the current node
856 * \param candidate_id the possible new predecessor
858 static void notify(node_t node, int predecessor_candidate_id) {
860 if (node->pred_id == -1
861 || is_in_interval(predecessor_candidate_id, node->pred_id + 1, node->id - 1)) {
863 set_predecessor(node, predecessor_candidate_id);
864 print_finger_table(node);
867 XBT_DEBUG("I don't have to change my predecessor to %d", predecessor_candidate_id);
872 * \brief Notifies a remote node that its predecessor may have changed.
873 * \param node the current node
874 * \param notify_id id of the node to notify
875 * \param candidate_id the possible new predecessor
877 static void remote_notify(node_t node, int notify_id, int predecessor_candidate_id) {
879 task_data_t req_data = xbt_new0(s_task_data_t, 1);
880 req_data->type = TASK_NOTIFY;
881 req_data->request_id = predecessor_candidate_id;
882 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
884 // send a "Notify" request to notify_id
885 msg_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
886 XBT_DEBUG("Sending a 'Notify' request (task %p) to %d", task, notify_id);
887 char mailbox[MAILBOX_NAME_SIZE];
888 get_mailbox(notify_id, mailbox);
889 MSG_task_dsend(task, mailbox, task_free);
893 * \brief This function is called periodically.
894 * It refreshes the finger table of the current node.
895 * \param node the current node
897 static void fix_fingers(node_t node) {
899 XBT_DEBUG("Fixing fingers");
900 int i = node->next_finger_to_fix;
901 int id = find_successor(node, node->id + powers2[i]);
904 if (id != node->fingers[i].id) {
905 set_finger(node, i, id);
906 print_finger_table(node);
908 node->next_finger_to_fix = (i + 1) % nb_bits;
913 * \brief This function is called periodically.
914 * It checks whether the predecessor has failed
915 * \param node the current node
917 static void check_predecessor(node_t node)
919 XBT_DEBUG("Checking whether my predecessor is alive");
921 if(node->pred_id == -1)
926 char mailbox[MAILBOX_NAME_SIZE];
927 get_mailbox(node->pred_id, mailbox);
928 task_data_t req_data = xbt_new0(s_task_data_t,1);
929 req_data->type = TASK_PREDECESSOR_ALIVE;
930 req_data->request_id = node->pred_id;
931 get_mailbox(node->id, req_data->answer_to);
932 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
934 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
935 XBT_DEBUG("Sending a 'Predecessor Alive' request to my predecessor %d", node->pred_id);
937 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
940 XBT_DEBUG("Failed to send the 'Predecessor Alive' request (task %p) to %d", task_sent, node->pred_id);
941 task_free(task_sent);
944 // receive the answer
945 XBT_DEBUG("Sent 'Predecessor Alive' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
946 task_sent, node->pred_id, req_data->answer_to);
949 if (node->comm_receive == NULL) { // FIXME simplify this
950 msg_task_t task_received = NULL;
951 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
954 res = MSG_comm_wait(node->comm_receive, timeout);
957 XBT_DEBUG("Failed to receive the answer to my 'Predecessor Alive' request (task %p): %d",
958 task_sent, (int)res);
960 MSG_comm_destroy(node->comm_receive);
961 node->comm_receive = NULL;
964 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
965 if (task_received != task_sent) {
966 MSG_comm_destroy(node->comm_receive);
967 node->comm_receive = NULL;
968 handle_task(node, task_received);
970 XBT_DEBUG("Received the answer to my 'Predecessor Alive' request (task %p) : my predecessor %d is alive", task_received, node->pred_id);
972 MSG_comm_destroy(node->comm_receive);
973 node->comm_receive = NULL;
974 task_free(task_received);
982 * \brief Performs a find successor request to a random id.
983 * \param node the current node
985 static void random_lookup(node_t node)
989 find_successor(node, id);
991 /*** Random lookup disabled for tesh examples ***/
992 /*if(node->stream == NULL)
993 node->stream = RngStream_CreateStream("");
994 int random_index = RngStream_RandInt (node->stream, 0, nb_bits - 1);
995 int random_id = node->fingers[random_index].id;
996 XBT_DEBUG("Making a lookup request for id %d", random_id);
997 int res = find_successor(node, random_id);
998 XBT_DEBUG("The successor of node %d is %d", random_id, res);*/
1003 * \brief Main function.
1005 int main(int argc, char *argv[])
1007 MSG_init(&argc, argv);
1009 printf("Usage: %s [-nb_bits=n] [-timeout=t] platform_file deployment_file\n", argv[0]);
1010 printf("example: %s ../msg_platform.xml chord.xml\n", argv[0]);
1014 char **options = &argv[1];
1015 while (!strncmp(options[0], "-", 1)) {
1017 int length = strlen("-nb_bits=");
1018 if (!strncmp(options[0], "-nb_bits=", length) && strlen(options[0]) > length) {
1019 nb_bits = atoi(options[0] + length);
1020 XBT_DEBUG("Set nb_bits to %d", nb_bits);
1024 length = strlen("-timeout=");
1025 if (!strncmp(options[0], "-timeout=", length) && strlen(options[0]) > length) {
1026 timeout = atoi(options[0] + length);
1027 XBT_DEBUG("Set timeout to %d", timeout);
1030 xbt_die("Invalid chord option '%s'", options[0]);
1036 const char* platform_file = options[0];
1037 const char* application_file = options[1];
1041 MSG_create_environment(platform_file);
1043 MSG_function_register("node", node);
1044 MSG_launch_application(application_file);
1046 msg_error_t res = MSG_main();
1047 XBT_CRITICAL("Messages created: %ld", smx_total_comms);
1048 XBT_INFO("Simulated time: %g", MSG_get_clock());