1 /* Copyright (c) 2010-2014. The SimGrid Team.
2 * All rights reserved. */
4 /* This program is free software; you can redistribute it and/or modify it
5 * under the terms of the license (GNU LGPL) which comes with this package. */
8 #include "simgrid/msg.h"
10 #include "xbt/asserts.h"
11 #include "simgrid/modelchecker.h"
12 #include <xbt/RngStream.h>
13 #include "mc/mc_replay.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 static const double sleep_delay = 4.9999;
43 extern long int smx_total_comms;
48 typedef struct s_finger {
50 char mailbox[MAILBOX_NAME_SIZE]; // string representation of the id
51 } s_finger_t, *finger_t;
56 typedef struct s_node {
58 char mailbox[MAILBOX_NAME_SIZE]; // my mailbox name (string representation of the id)
59 s_finger_t *fingers; // finger table, of size nb_bits (fingers[0] is my successor)
60 int pred_id; // predecessor id
61 char pred_mailbox[MAILBOX_NAME_SIZE]; // predecessor's mailbox name
62 int next_finger_to_fix; // index of the next finger to fix in fix_fingers()
63 msg_comm_t comm_receive; // current communication to receive
64 double last_change_date; // last time I changed a finger or my predecessor
65 RngStream stream; //RngStream for
69 * Types of tasks exchanged between nodes.
73 TASK_FIND_SUCCESSOR_ANSWER,
75 TASK_GET_PREDECESSOR_ANSWER,
77 TASK_SUCCESSOR_LEAVING,
78 TASK_PREDECESSOR_LEAVING,
79 TASK_PREDECESSOR_ALIVE,
80 TASK_PREDECESSOR_ALIVE_ANSWER
84 * Data attached with the tasks sent and received
86 typedef struct s_task_data {
87 e_task_type_t type; // type of task
88 int request_id; // id paramater (used by some types of tasks)
89 int request_finger; // finger parameter (used by some types of tasks)
90 int answer_id; // answer (used by some types of tasks)
91 char answer_to[MAILBOX_NAME_SIZE]; // mailbox to send an answer to (if any)
92 const char* issuer_host_name; // used for logging
93 } s_task_data_t, *task_data_t;
96 static xbt_dynar_t host_list;
99 static void chord_initialize(void);
100 static void chord_exit(void);
101 static int normalize(int id);
102 static int is_in_interval(int id, int start, int end);
103 static void get_mailbox(int host_id, char* mailbox);
104 static void task_free(void* task);
105 static void print_finger_table(node_t node);
106 static void set_finger(node_t node, int finger_index, int id);
107 static void set_predecessor(node_t node, int predecessor_id);
110 static int node(int argc, char *argv[]);
111 static void handle_task(node_t node, msg_task_t task);
114 static void create(node_t node);
115 static int join(node_t node, int known_id);
116 static void leave(node_t node);
117 static int find_successor(node_t node, int id);
118 static int remote_find_successor(node_t node, int ask_to_id, int id);
119 static int remote_get_predecessor(node_t node, int ask_to_id);
120 static int closest_preceding_node(node_t node, int id);
121 static void stabilize(node_t node);
122 static void notify(node_t node, int predecessor_candidate_id);
123 static void remote_notify(node_t node, int notify_to, int predecessor_candidate_id);
124 static void fix_fingers(node_t node);
125 static void check_predecessor(node_t node);
126 static void random_lookup(node_t);
127 static void quit_notify(node_t node);
130 * \brief Global initialization of the Chord simulation.
132 static void chord_initialize(void)
134 // compute the powers of 2 once for all
135 powers2 = xbt_new(int, nb_bits);
138 for (i = 0; i < nb_bits; i++) {
143 XBT_DEBUG("Sets nb_keys to %d", nb_keys);
146 host_list = MSG_hosts_as_dynar();
147 xbt_dynar_foreach(host_list, i, host) {
150 snprintf(descr, sizeof descr, "RngSream<%s>", MSG_host_get_name(host));
151 stream = RngStream_CreateStream(descr);
152 MSG_host_set_property_value(host, "stream", (char*)stream, NULL);
156 static void chord_exit(void)
160 xbt_dynar_foreach(host_list, i, host) {
161 RngStream stream = (RngStream)MSG_host_get_property_value(host, "stream");
162 RngStream_DeleteStream(&stream);
164 xbt_dynar_free(&host_list);
170 * \brief Turns an id into an equivalent id in [0, nb_keys).
172 * \return the corresponding normalized id
174 static int normalize(int id)
176 // like id % nb_keys, but works with negatives numbers (and faster)
177 return id & (nb_keys - 1);
181 * \brief Returns whether an id belongs to the interval [start, end].
183 * The parameters are noramlized to make sure they are between 0 and nb_keys - 1).
184 * 1 belongs to [62, 3]
185 * 1 does not belong to [3, 62]
186 * 63 belongs to [62, 3]
187 * 63 does not belong to [3, 62]
188 * 24 belongs to [21, 29]
189 * 24 does not belong to [29, 21]
191 * \param id id to check
192 * \param start lower bound
193 * \param end upper bound
194 * \return a non-zero value if id in in [start, end]
196 static int is_in_interval(int id, int start, int end)
199 start = normalize(start);
200 end = normalize(end);
202 // make sure end >= start and id >= start
215 * \brief Gets the mailbox name of a host given its chord id.
216 * \param node_id id of a node
217 * \param mailbox pointer to where the mailbox name should be written
218 * (there must be enough space)
220 static void get_mailbox(int node_id, char* mailbox)
222 snprintf(mailbox, MAILBOX_NAME_SIZE - 1, "%d", node_id);
226 * \brief Frees the memory used by a task.
227 * \param task the MSG task to destroy
229 static void task_free(void* task)
231 // TODO add a parameter data_free_function to MSG_task_create?
233 xbt_free(MSG_task_get_data(task));
234 MSG_task_destroy(task);
239 * \brief Displays the finger table of a node.
242 static void print_finger_table(node_t node)
244 if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) {
246 XBT_VERB("My finger table:");
247 XBT_VERB("Start | Succ");
248 for (i = 0; i < nb_bits; i++) {
249 XBT_VERB(" %3d | %3d", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
251 XBT_VERB("Predecessor: %d", node->pred_id);
256 * \brief Sets a finger of the current node.
257 * \param node the current node
258 * \param finger_index index of the finger to set (0 to nb_bits - 1)
259 * \param id the id to set for this finger
261 static void set_finger(node_t node, int finger_index, int id)
263 if (id != node->fingers[finger_index].id) {
264 node->fingers[finger_index].id = id;
265 get_mailbox(id, node->fingers[finger_index].mailbox);
266 node->last_change_date = MSG_get_clock();
267 XBT_DEBUG("My new finger #%d is %d", finger_index, id);
272 * \brief Sets the predecessor of the current node.
273 * \param node the current node
274 * \param id the id to predecessor, or -1 to unset the predecessor
276 static void set_predecessor(node_t node, int predecessor_id)
278 if (predecessor_id != node->pred_id) {
279 node->pred_id = predecessor_id;
281 if (predecessor_id != -1) {
282 get_mailbox(predecessor_id, node->pred_mailbox);
284 node->last_change_date = MSG_get_clock();
286 XBT_DEBUG("My new predecessor is %d", predecessor_id);
291 * \brief Node Function
294 * - the id of a guy I know in the system (except for the first node)
295 * - the time to sleep before I join (except for the first node)
297 int node(int argc, char *argv[])
300 /* Reduce the run size for the MC */
301 if(MC_is_active() || MC_record_replay_is_active()){
302 periodic_stabilize_delay = 8;
303 periodic_fix_fingers_delay = 8;
304 periodic_check_predecessor_delay = 8;
307 double init_time = MSG_get_clock();
308 msg_task_t task_received = NULL;
310 int join_success = 0;
312 double next_stabilize_date = init_time + periodic_stabilize_delay;
313 double next_fix_fingers_date = init_time + periodic_fix_fingers_delay;
314 double next_check_predecessor_date = init_time + periodic_check_predecessor_delay;
315 double next_lookup_date = init_time + periodic_lookup_delay;
319 int sub_protocol = 0;
321 xbt_assert(argc == 3 || argc == 5, "Wrong number of arguments for this node");
323 // initialize my node
325 node.id = atoi(argv[1]);
327 (RngStream)MSG_host_get_property_value(MSG_host_self(), "stream");
328 get_mailbox(node.id, node.mailbox);
329 node.next_finger_to_fix = 0;
330 node.fingers = xbt_new0(s_finger_t, nb_bits);
331 node.last_change_date = init_time;
333 for (i = 0; i < nb_bits; i++) {
334 node.fingers[i].id = -1;
335 set_finger(&node, i, node.id);
338 if (argc == 3) { // first ring
339 deadline = atof(argv[2]);
344 int known_id = atoi(argv[2]);
345 //double sleep_time = atof(argv[3]);
346 deadline = atof(argv[4]);
349 // sleep before starting
350 XBT_DEBUG("Let's sleep during %f", sleep_time);
351 MSG_process_sleep(sleep_time);
353 XBT_DEBUG("Hey! Let's join the system.");
355 join_success = join(&node, known_id);
359 while (MSG_get_clock() < init_time + deadline
360 // && MSG_get_clock() < node.last_change_date + 1000
361 && MSG_get_clock() < max_simulation_time) {
363 if (node.comm_receive == NULL) {
364 task_received = NULL;
365 node.comm_receive = MSG_task_irecv(&task_received, node.mailbox);
366 // FIXME: do not make MSG_task_irecv() calls from several functions
369 //XBT_INFO("Node %d is ring member : %d", node.id, is_ring_member(known_id, node.id) != -1);
371 if (!MSG_comm_test(node.comm_receive)) {
373 // no task was received: make some periodic calls
375 if(MC_is_active() || MC_record_replay_is_active()){
376 if(MC_is_active() && !MC_visited_reduction() && no_op){
379 if(listen == 0 && (sub_protocol = MC_random(0, 4)) > 0){
380 if(sub_protocol == 1)
382 else if(sub_protocol == 2)
384 else if(sub_protocol == 3)
385 check_predecessor(&node);
387 random_lookup(&node);
390 MSG_process_sleep(sleep_delay);
391 if(!MC_visited_reduction())
395 if (MSG_get_clock() >= next_stabilize_date) {
397 next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay;
398 }else if (MSG_get_clock() >= next_fix_fingers_date) {
400 next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay;
401 }else if (MSG_get_clock() >= next_check_predecessor_date) {
402 check_predecessor(&node);
403 next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay;
404 }else if (MSG_get_clock() >= next_lookup_date) {
405 random_lookup(&node);
406 next_lookup_date = MSG_get_clock() + periodic_lookup_delay;
408 // nothing to do: sleep for a while
409 MSG_process_sleep(sleep_delay);
414 // a transfer has occurred
416 msg_error_t status = MSG_comm_get_status(node.comm_receive);
418 if (status != MSG_OK) {
419 XBT_DEBUG("Failed to receive a task. Nevermind.");
420 MSG_comm_destroy(node.comm_receive);
421 node.comm_receive = NULL;
424 // the task was successfully received
425 MSG_comm_destroy(node.comm_receive);
426 node.comm_receive = NULL;
427 handle_task(&node, task_received);
432 if (node.comm_receive) {
433 /* handle last task if any */
434 if (MSG_comm_wait(node.comm_receive, 0) == MSG_OK)
435 task_free(task_received);
436 MSG_comm_destroy(node.comm_receive);
437 node.comm_receive = NULL;
444 // stop the simulation
445 xbt_free(node.fingers);
450 * \brief This function is called when the current node receives a task.
451 * \param node the current node
452 * \param task the task to handle (don't touch it then:
453 * it will be destroyed, reused or forwarded)
455 static void handle_task(node_t node, msg_task_t task) {
457 XBT_DEBUG("Handling task %p", task);
458 char mailbox[MAILBOX_NAME_SIZE];
459 task_data_t task_data = (task_data_t) MSG_task_get_data(task);
460 e_task_type_t type = task_data->type;
464 case TASK_FIND_SUCCESSOR:
465 XBT_DEBUG("Receiving a 'Find Successor' request from %s for id %d",
466 task_data->issuer_host_name, task_data->request_id);
467 // is my successor the successor?
468 if (is_in_interval(task_data->request_id, node->id + 1, node->fingers[0].id)) {
469 task_data->type = TASK_FIND_SUCCESSOR_ANSWER;
470 task_data->answer_id = node->fingers[0].id;
471 XBT_DEBUG("Sending back a 'Find Successor Answer' to %s (mailbox %s): the successor of %d is %d",
472 task_data->issuer_host_name,
473 task_data->answer_to,
474 task_data->request_id, task_data->answer_id);
475 MSG_task_dsend(task, task_data->answer_to, task_free);
478 // otherwise, forward the request to the closest preceding finger in my table
479 int closest = closest_preceding_node(node, task_data->request_id);
480 XBT_DEBUG("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d",
481 task_data->request_id, closest);
482 get_mailbox(closest, mailbox);
483 MSG_task_dsend(task, mailbox, task_free);
487 case TASK_GET_PREDECESSOR:
488 XBT_DEBUG("Receiving a 'Get Predecessor' request from %s", task_data->issuer_host_name);
489 task_data->type = TASK_GET_PREDECESSOR_ANSWER;
490 task_data->answer_id = node->pred_id;
491 XBT_DEBUG("Sending back a 'Get Predecessor Answer' to %s via mailbox '%s': my predecessor is %d",
492 task_data->issuer_host_name,
493 task_data->answer_to, task_data->answer_id);
494 MSG_task_dsend(task, task_data->answer_to, task_free);
498 // someone is telling me that he may be my new predecessor
499 XBT_DEBUG("Receiving a 'Notify' request from %s", task_data->issuer_host_name);
500 notify(node, task_data->request_id);
504 case TASK_PREDECESSOR_LEAVING:
505 // my predecessor is about to quit
506 XBT_DEBUG("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name);
507 // modify my predecessor
508 set_predecessor(node, task_data->request_id);
511 >> notify my new predecessor
512 >> send a notify_predecessors !!
516 case TASK_SUCCESSOR_LEAVING:
517 // my successor is about to quit
518 XBT_DEBUG("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name);
519 // modify my successor FIXME : this should be implicit ?
520 set_finger(node, 0, task_data->request_id);
523 >> notify my new successor
524 >> update my table & predecessors table */
527 case TASK_FIND_SUCCESSOR_ANSWER:
528 case TASK_GET_PREDECESSOR_ANSWER:
529 case TASK_PREDECESSOR_ALIVE_ANSWER:
530 XBT_DEBUG("Ignoring unexpected task of type %d (%p)", (int)type, task);
534 case TASK_PREDECESSOR_ALIVE:
535 XBT_DEBUG("Receiving a 'Predecessor Alive' request from %s", task_data->issuer_host_name);
536 task_data->type = TASK_PREDECESSOR_ALIVE_ANSWER;
537 XBT_DEBUG("Sending back a 'Predecessor Alive Answer' to %s (mailbox %s)",
538 task_data->issuer_host_name,
539 task_data->answer_to);
540 MSG_task_dsend(task, task_data->answer_to, task_free);
549 * \brief Initializes the current node as the first one of the system.
550 * \param node the current node
552 static void create(node_t node)
554 XBT_DEBUG("Create a new Chord ring...");
555 set_predecessor(node, -1); // -1 means that I have no predecessor
556 print_finger_table(node);
560 * \brief Makes the current node join the ring, knowing the id of a node
561 * already in the ring
562 * \param node the current node
563 * \param known_id id of a node already in the ring
564 * \return 1 if the join operation succeeded, 0 otherwise
566 static int join(node_t node, int known_id)
568 XBT_INFO("Joining the ring with id %d, knowing node %d", node->id, known_id);
569 set_predecessor(node, -1); // no predecessor (yet)
573 for (i = 0; i < nb_bits; i++) {
574 set_finger(node, i, known_id);
578 int successor_id = remote_find_successor(node, known_id, node->id);
579 if (successor_id == -1) {
580 XBT_INFO("Cannot join the ring.");
583 set_finger(node, 0, successor_id);
584 print_finger_table(node);
587 return successor_id != -1;
591 * \brief Makes the current node quit the system
592 * \param node the current node
594 static void leave(node_t node)
596 XBT_DEBUG("Well Guys! I Think it's time for me to quit ;)");
601 * \brief Notifies the successor and the predecessor of the current node
603 * \param node the current node
605 static void quit_notify(node_t node)
607 char mailbox[MAILBOX_NAME_SIZE];
608 //send the PREDECESSOR_LEAVING to our successor
609 task_data_t req_data = xbt_new0(s_task_data_t,1);
610 req_data->type = TASK_PREDECESSOR_LEAVING;
611 req_data->request_id = node->pred_id;
612 get_mailbox(node->id, req_data->answer_to);
613 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
615 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
616 XBT_DEBUG("Sending a 'PREDECESSOR_LEAVING' to my successor %d",node->fingers[0].id);
617 if (MSG_task_send_with_timeout(task_sent, node->fingers[0].mailbox, timeout)==
619 XBT_DEBUG("Timeout expired when sending a 'PREDECESSOR_LEAVING' to my successor %d",
620 node->fingers[0].id);
621 task_free(task_sent);
624 //send the SUCCESSOR_LEAVING to our predecessor
625 get_mailbox(node->pred_id, mailbox);
626 task_data_t req_data_s = xbt_new0(s_task_data_t,1);
627 req_data_s->type = TASK_SUCCESSOR_LEAVING;
628 req_data_s->request_id = node->fingers[0].id;
629 req_data_s->request_id = node->pred_id;
630 get_mailbox(node->id, req_data_s->answer_to);
631 req_data_s->issuer_host_name = MSG_host_get_name(MSG_host_self());
633 msg_task_t task_sent_s = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data_s);
634 XBT_DEBUG("Sending a 'SUCCESSOR_LEAVING' to my predecessor %d",node->pred_id);
635 if (MSG_task_send_with_timeout(task_sent_s, mailbox, timeout)==
637 XBT_DEBUG("Timeout expired when sending a 'SUCCESSOR_LEAVING' to my predecessor %d",
639 task_free(task_sent_s);
645 * \brief Makes the current node find the successor node of an id.
646 * \param node the current node
647 * \param id the id to find
648 * \return the id of the successor node, or -1 if the request failed
650 static int find_successor(node_t node, int id)
652 // is my successor the successor?
653 if (is_in_interval(id, node->id + 1, node->fingers[0].id)) {
654 return node->fingers[0].id;
657 // otherwise, ask the closest preceding finger in my table
658 int closest = closest_preceding_node(node, id);
659 return remote_find_successor(node, closest, id);
663 * \brief Asks another node the successor node of an id.
664 * \param node the current node
665 * \param ask_to the node to ask to
666 * \param id the id to find
667 * \return the id of the successor node, or -1 if the request failed
669 static int remote_find_successor(node_t node, int ask_to, int id)
673 char mailbox[MAILBOX_NAME_SIZE];
674 get_mailbox(ask_to, mailbox);
675 task_data_t req_data = xbt_new0(s_task_data_t, 1);
676 req_data->type = TASK_FIND_SUCCESSOR;
677 req_data->request_id = id;
678 get_mailbox(node->id, req_data->answer_to);
679 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
681 // send a "Find Successor" request to ask_to_id
682 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
683 XBT_DEBUG("Sending a 'Find Successor' request (task %p) to %d for id %d", task_sent, ask_to, id);
684 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
687 XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
688 task_sent, ask_to, id);
689 task_free(task_sent);
693 // receive the answer
694 XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
695 task_sent, ask_to, id);
698 if (node->comm_receive == NULL) {
699 msg_task_t task_received = NULL;
700 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
703 res = MSG_comm_wait(node->comm_receive, timeout);
706 XBT_DEBUG("Failed to receive the answer to my 'Find Successor' request (task %p): %d",
707 task_sent, (int)res);
709 MSG_comm_destroy(node->comm_receive);
710 node->comm_receive = NULL;
713 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
714 XBT_DEBUG("Received a task (%p)", task_received);
715 task_data_t ans_data = MSG_task_get_data(task_received);
717 // Once upon a time, our code assumed that here, task_received != task_sent all the time
719 // This assumption is wrong (as messages from differing round can interleave), leading to a bug in our code.
720 // We failed to find this bug directly, as it only occured on large platforms, leading to hardly usable traces.
721 // Instead, we used the model-checker to track down the issue by adding the following test here in the code:
722 // if (MC_is_active()) {
723 // MC_assert(task_received == task_sent);
725 // That explained the bug in a snap, with a very cool example and everything.
727 // This MC_assert is now desactivated as the case is now properly handled in our code and we don't want the
728 // MC to fail any further under that condition, but this comment is here to as a memorial for this first
729 // brillant victory of the model-checking in the SimGrid community :)
731 if (task_received != task_sent ||
732 ans_data->type != TASK_FIND_SUCCESSOR_ANSWER) {
733 // this is not the expected answer
734 MSG_comm_destroy(node->comm_receive);
735 node->comm_receive = NULL;
736 handle_task(node, task_received);
739 // this is our answer
740 XBT_DEBUG("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d",
741 ans_data->request_id, task_received, id, ans_data->answer_id);
742 successor = ans_data->answer_id;
744 MSG_comm_destroy(node->comm_receive);
745 node->comm_receive = NULL;
746 task_free(task_received);
756 * \brief Asks another node its predecessor.
757 * \param node the current node
758 * \param ask_to the node to ask to
759 * \return the id of its predecessor node, or -1 if the request failed
760 * (or if the node does not know its predecessor)
762 static int remote_get_predecessor(node_t node, int ask_to)
764 int predecessor_id = -1;
766 char mailbox[MAILBOX_NAME_SIZE];
767 get_mailbox(ask_to, mailbox);
768 task_data_t req_data = xbt_new0(s_task_data_t, 1);
769 req_data->type = TASK_GET_PREDECESSOR;
770 get_mailbox(node->id, req_data->answer_to);
771 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
773 // send a "Get Predecessor" request to ask_to_id
774 XBT_DEBUG("Sending a 'Get Predecessor' request to %d", ask_to);
775 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
776 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
779 XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d",
781 task_free(task_sent);
785 // receive the answer
786 XBT_DEBUG("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
787 task_sent, ask_to, req_data->answer_to);
790 if (node->comm_receive == NULL) { // FIXME simplify this
791 msg_task_t task_received = NULL;
792 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
795 res = MSG_comm_wait(node->comm_receive, timeout);
798 XBT_DEBUG("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d",
799 task_sent, (int)res);
801 MSG_comm_destroy(node->comm_receive);
802 node->comm_receive = NULL;
805 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
806 task_data_t ans_data = MSG_task_get_data(task_received);
808 /*if (MC_is_active()) {
809 MC_assert(task_received == task_sent);
812 if (task_received != task_sent ||
813 ans_data->type != TASK_GET_PREDECESSOR_ANSWER) {
814 MSG_comm_destroy(node->comm_receive);
815 node->comm_receive = NULL;
816 handle_task(node, task_received);
819 XBT_DEBUG("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d",
820 task_received, ask_to, ans_data->answer_id);
821 predecessor_id = ans_data->answer_id;
823 MSG_comm_destroy(node->comm_receive);
824 node->comm_receive = NULL;
825 task_free(task_received);
831 return predecessor_id;
835 * \brief Returns the closest preceding finger of an id
836 * with respect to the finger table of the current node.
837 * \param node the current node
838 * \param id the id to find
839 * \return the closest preceding finger of that id
841 int closest_preceding_node(node_t node, int id)
844 for (i = nb_bits - 1; i >= 0; i--) {
845 if (is_in_interval(node->fingers[i].id, node->id + 1, id - 1)) {
846 return node->fingers[i].id;
853 * \brief This function is called periodically. It checks the immediate
854 * successor of the current node.
855 * \param node the current node
857 static void stabilize(node_t node)
859 XBT_DEBUG("Stabilizing node");
861 // get the predecessor of my immediate successor
863 int successor_id = node->fingers[0].id;
864 if (successor_id != node->id) {
865 candidate_id = remote_get_predecessor(node, successor_id);
868 candidate_id = node->pred_id;
871 // this node is a candidate to become my new successor
872 if (candidate_id != -1
873 && is_in_interval(candidate_id, node->id + 1, successor_id - 1)) {
874 set_finger(node, 0, candidate_id);
876 if (successor_id != node->id) {
877 remote_notify(node, successor_id, node->id);
882 * \brief Notifies the current node that its predecessor may have changed.
883 * \param node the current node
884 * \param candidate_id the possible new predecessor
886 static void notify(node_t node, int predecessor_candidate_id) {
888 if (node->pred_id == -1
889 || is_in_interval(predecessor_candidate_id, node->pred_id + 1, node->id - 1)) {
891 set_predecessor(node, predecessor_candidate_id);
892 print_finger_table(node);
895 XBT_DEBUG("I don't have to change my predecessor to %d", predecessor_candidate_id);
900 * \brief Notifies a remote node that its predecessor may have changed.
901 * \param node the current node
902 * \param notify_id id of the node to notify
903 * \param candidate_id the possible new predecessor
905 static void remote_notify(node_t node, int notify_id, int predecessor_candidate_id) {
907 task_data_t req_data = xbt_new0(s_task_data_t, 1);
908 req_data->type = TASK_NOTIFY;
909 req_data->request_id = predecessor_candidate_id;
910 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
912 // send a "Notify" request to notify_id
913 msg_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
914 XBT_DEBUG("Sending a 'Notify' request (task %p) to %d", task, notify_id);
915 char mailbox[MAILBOX_NAME_SIZE];
916 get_mailbox(notify_id, mailbox);
917 MSG_task_dsend(task, mailbox, task_free);
921 * \brief This function is called periodically.
922 * It refreshes the finger table of the current node.
923 * \param node the current node
925 static void fix_fingers(node_t node) {
927 XBT_DEBUG("Fixing fingers");
928 int i = node->next_finger_to_fix;
929 int id = find_successor(node, node->id + powers2[i]);
932 if (id != node->fingers[i].id) {
933 set_finger(node, i, id);
934 print_finger_table(node);
936 node->next_finger_to_fix = (i + 1) % nb_bits;
941 * \brief This function is called periodically.
942 * It checks whether the predecessor has failed
943 * \param node the current node
945 static void check_predecessor(node_t node)
947 XBT_DEBUG("Checking whether my predecessor is alive");
949 if(node->pred_id == -1)
954 char mailbox[MAILBOX_NAME_SIZE];
955 get_mailbox(node->pred_id, mailbox);
956 task_data_t req_data = xbt_new0(s_task_data_t,1);
957 req_data->type = TASK_PREDECESSOR_ALIVE;
958 req_data->request_id = node->pred_id;
959 get_mailbox(node->id, req_data->answer_to);
960 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
962 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
963 XBT_DEBUG("Sending a 'Predecessor Alive' request to my predecessor %d", node->pred_id);
965 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
968 XBT_DEBUG("Failed to send the 'Predecessor Alive' request (task %p) to %d", task_sent, node->pred_id);
969 task_free(task_sent);
972 // receive the answer
973 XBT_DEBUG("Sent 'Predecessor Alive' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
974 task_sent, node->pred_id, req_data->answer_to);
977 if (node->comm_receive == NULL) { // FIXME simplify this
978 msg_task_t task_received = NULL;
979 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
982 res = MSG_comm_wait(node->comm_receive, timeout);
985 XBT_DEBUG("Failed to receive the answer to my 'Predecessor Alive' request (task %p): %d",
986 task_sent, (int)res);
988 MSG_comm_destroy(node->comm_receive);
989 node->comm_receive = NULL;
992 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
993 if (task_received != task_sent) {
994 MSG_comm_destroy(node->comm_receive);
995 node->comm_receive = NULL;
996 handle_task(node, task_received);
998 XBT_DEBUG("Received the answer to my 'Predecessor Alive' request (task %p) : my predecessor %d is alive", task_received, node->pred_id);
1000 MSG_comm_destroy(node->comm_receive);
1001 node->comm_receive = NULL;
1002 task_free(task_received);
1010 * \brief Performs a find successor request to a random id.
1011 * \param node the current node
1013 static void random_lookup(node_t node)
1015 int random_index = RngStream_RandInt (node->stream, 0, nb_bits - 1);
1016 int random_id = node->fingers[random_index].id;
1017 XBT_DEBUG("Making a lookup request for id %d", random_id);
1018 int res = find_successor(node, random_id);
1019 XBT_DEBUG("The successor of node %d is %d", random_id, res);
1024 * \brief Main function.
1026 int main(int argc, char *argv[])
1028 MSG_init(&argc, argv);
1030 printf("Usage: %s [-nb_bits=n] [-timeout=t] platform_file deployment_file\n", argv[0]);
1031 printf("example: %s ../msg_platform.xml chord.xml\n", argv[0]);
1035 char **options = &argv[1];
1036 while (!strncmp(options[0], "-", 1)) {
1038 int length = strlen("-nb_bits=");
1039 if (!strncmp(options[0], "-nb_bits=", length) && strlen(options[0]) > length) {
1040 nb_bits = atoi(options[0] + length);
1041 XBT_DEBUG("Set nb_bits to %d", nb_bits);
1045 length = strlen("-timeout=");
1046 if (!strncmp(options[0], "-timeout=", length) && strlen(options[0]) > length) {
1047 timeout = atoi(options[0] + length);
1048 XBT_DEBUG("Set timeout to %d", timeout);
1051 xbt_die("Invalid chord option '%s'", options[0]);
1057 const char* platform_file = options[0];
1058 const char* application_file = options[1];
1060 MSG_create_environment(platform_file);
1064 MSG_function_register("node", node);
1065 MSG_launch_application(application_file);
1067 msg_error_t res = MSG_main();
1068 XBT_CRITICAL("Messages created: %ld", smx_total_comms);
1069 XBT_INFO("Simulated time: %g", MSG_get_clock());