1 /* Copyright (c) 2010-2015. 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. */
7 #include "simgrid/msg.h"
8 #include "simgrid/modelchecker.h"
9 #include <xbt/RngStream.h>
10 #include "src/mc/mc_replay.h" // FIXME: this is an internal header
12 /** @addtogroup MSG_examples
14 * - <b>Chord P2P protocol dht-chord/dht-chord.c:</b>. This example implements the well known Chord P2P protocol. Its
15 * main advantage is that it constitutes a fully working non-trivial example. In addition, its implementation is
16 * rather efficient, as demonstrated in http://hal.inria.fr/inria-00602216/
20 XBT_LOG_NEW_DEFAULT_CATEGORY(msg_chord, "Messages specific for this msg example");
24 #define MAILBOX_NAME_SIZE 10
26 static int nb_bits = 24;
27 static int nb_keys = 0;
28 static int timeout = 50;
29 static int max_simulation_time = 1000;
30 static int periodic_stabilize_delay = 20;
31 static int periodic_fix_fingers_delay = 120;
32 static int periodic_check_predecessor_delay = 120;
33 static int periodic_lookup_delay = 10;
35 static const double sleep_delay = 4.9999;
37 extern long int smx_total_comms;
40 typedef struct s_finger {
42 char mailbox[MAILBOX_NAME_SIZE]; // string representation of the id
43 } s_finger_t, *finger_t;
46 typedef struct s_node {
48 char mailbox[MAILBOX_NAME_SIZE]; // my mailbox name (string representation of the id)
49 s_finger_t *fingers; // finger table, of size nb_bits (fingers[0] is my successor)
50 int pred_id; // predecessor id
51 char pred_mailbox[MAILBOX_NAME_SIZE]; // predecessor's mailbox name
52 int next_finger_to_fix; // index of the next finger to fix in fix_fingers()
53 msg_comm_t comm_receive; // current communication to receive
54 double last_change_date; // last time I changed a finger or my predecessor
55 RngStream stream; //RngStream for
58 /* Types of tasks exchanged between nodes. */
61 TASK_FIND_SUCCESSOR_ANSWER,
63 TASK_GET_PREDECESSOR_ANSWER,
65 TASK_SUCCESSOR_LEAVING,
66 TASK_PREDECESSOR_LEAVING,
67 TASK_PREDECESSOR_ALIVE,
68 TASK_PREDECESSOR_ALIVE_ANSWER
71 /* Data attached with the tasks sent and received */
72 typedef struct s_task_data {
73 e_task_type_t type; // type of task
74 int request_id; // id paramater (used by some types of tasks)
75 int request_finger; // finger parameter (used by some types of tasks)
76 int answer_id; // answer (used by some types of tasks)
77 char answer_to[MAILBOX_NAME_SIZE]; // mailbox to send an answer to (if any)
78 const char* issuer_host_name; // used for logging
79 } s_task_data_t, *task_data_t;
82 static xbt_dynar_t host_list;
85 static void chord_exit(void);
86 static int normalize(int id);
87 static int is_in_interval(int id, int start, int end);
88 static void get_mailbox(int host_id, char* mailbox);
89 static void task_free(void* task);
90 static void print_finger_table(node_t node);
91 static void set_finger(node_t node, int finger_index, int id);
92 static void set_predecessor(node_t node, int predecessor_id);
95 static int node(int argc, char *argv[]);
96 static void handle_task(node_t node, msg_task_t task);
99 static void create(node_t node);
100 static int join(node_t node, int known_id);
101 static void leave(node_t node);
102 static int find_successor(node_t node, int id);
103 static int remote_find_successor(node_t node, int ask_to_id, int id);
104 static int remote_get_predecessor(node_t node, int ask_to_id);
105 static int closest_preceding_node(node_t node, int id);
106 static void stabilize(node_t node);
107 static void notify(node_t node, int predecessor_candidate_id);
108 static void remote_notify(node_t node, int notify_to, int predecessor_candidate_id);
109 static void fix_fingers(node_t node);
110 static void check_predecessor(node_t node);
111 static void random_lookup(node_t);
112 static void quit_notify(node_t node);
114 /* \brief Global initialization of the Chord simulation. */
115 static void chord_initialize(void)
117 // compute the powers of 2 once for all
118 powers2 = xbt_new(int, nb_bits);
121 for (i = 0; i < nb_bits; i++) {
126 XBT_DEBUG("Sets nb_keys to %d", nb_keys);
129 host_list = MSG_hosts_as_dynar();
130 xbt_dynar_foreach(host_list, i, host) {
133 snprintf(descr, sizeof descr, "RngSream<%s>", MSG_host_get_name(host));
134 stream = RngStream_CreateStream(descr);
135 MSG_host_set_property_value(host, "stream", (char*)stream, NULL);
139 static void chord_exit(void)
143 xbt_dynar_foreach(host_list, i, host) {
144 RngStream stream = (RngStream)MSG_host_get_property_value(host, "stream");
145 RngStream_DeleteStream(&stream);
147 xbt_dynar_free(&host_list);
153 * \brief Turns an id into an equivalent id in [0, nb_keys).
155 * \return the corresponding normalized id
157 static int normalize(int id)
159 // like id % nb_keys, but works with negatives numbers (and faster)
160 return id & (nb_keys - 1);
164 * \brief Returns whether an id belongs to the interval [start, end].
166 * The parameters are noramlized to make sure they are between 0 and nb_keys - 1).
167 * 1 belongs to [62, 3]
168 * 1 does not belong to [3, 62]
169 * 63 belongs to [62, 3]
170 * 63 does not belong to [3, 62]
171 * 24 belongs to [21, 29]
172 * 24 does not belong to [29, 21]
174 * \param id id to check
175 * \param start lower bound
176 * \param end upper bound
177 * \return a non-zero value if id in in [start, end]
179 static int is_in_interval(int id, int start, int end)
182 start = normalize(start);
183 end = normalize(end);
185 // make sure end >= start and id >= start
198 * \brief Gets the mailbox name of a host given its chord id.
199 * \param node_id id of a node
200 * \param mailbox pointer to where the mailbox name should be written
201 * (there must be enough space)
203 static void get_mailbox(int node_id, char* mailbox)
205 snprintf(mailbox, MAILBOX_NAME_SIZE - 1, "%d", node_id);
209 * \brief Frees the memory used by a task.
210 * \param task the MSG task to destroy
212 static void task_free(void* task)
214 // TODO add a parameter data_free_function to MSG_task_create?
216 xbt_free(MSG_task_get_data(task));
217 MSG_task_destroy(task);
222 * \brief Displays the finger table of a node.
225 static void print_finger_table(node_t node)
227 if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) {
229 XBT_VERB("My finger table:");
230 XBT_VERB("Start | Succ");
231 for (i = 0; i < nb_bits; i++) {
232 XBT_VERB(" %3d | %3d", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
234 XBT_VERB("Predecessor: %d", node->pred_id);
239 * \brief Sets a finger of the current node.
240 * \param node the current node
241 * \param finger_index index of the finger to set (0 to nb_bits - 1)
242 * \param id the id to set for this finger
244 static void set_finger(node_t node, int finger_index, int id)
246 if (id != node->fingers[finger_index].id) {
247 node->fingers[finger_index].id = id;
248 get_mailbox(id, node->fingers[finger_index].mailbox);
249 node->last_change_date = MSG_get_clock();
250 XBT_DEBUG("My new finger #%d is %d", finger_index, id);
255 * \brief Sets the predecessor of the current node.
256 * \param node the current node
257 * \param id the id to predecessor, or -1 to unset the predecessor
259 static void set_predecessor(node_t node, int predecessor_id)
261 if (predecessor_id != node->pred_id) {
262 node->pred_id = predecessor_id;
264 if (predecessor_id != -1) {
265 get_mailbox(predecessor_id, node->pred_mailbox);
267 node->last_change_date = MSG_get_clock();
269 XBT_DEBUG("My new predecessor is %d", predecessor_id);
274 * \brief Node Function
277 * - the id of a guy I know in the system (except for the first node)
278 * - the time to sleep before I join (except for the first node)
280 int node(int argc, char *argv[])
283 /* Reduce the run size for the MC */
284 if(MC_is_active() || MC_record_replay_is_active()){
285 periodic_stabilize_delay = 8;
286 periodic_fix_fingers_delay = 8;
287 periodic_check_predecessor_delay = 8;
290 double init_time = MSG_get_clock();
291 msg_task_t task_received = NULL;
293 int join_success = 0;
295 double next_stabilize_date = init_time + periodic_stabilize_delay;
296 double next_fix_fingers_date = init_time + periodic_fix_fingers_delay;
297 double next_check_predecessor_date = init_time + periodic_check_predecessor_delay;
298 double next_lookup_date = init_time + periodic_lookup_delay;
302 int sub_protocol = 0;
304 xbt_assert(argc == 3 || argc == 5, "Wrong number of arguments for this node");
306 // initialize my node
308 node.id = xbt_str_parse_int(argv[1],"Invalid ID: %s");
309 node.stream = (RngStream)MSG_host_get_property_value(MSG_host_self(), "stream");
310 get_mailbox(node.id, node.mailbox);
311 node.next_finger_to_fix = 0;
312 node.fingers = xbt_new0(s_finger_t, nb_bits);
313 node.last_change_date = init_time;
315 for (i = 0; i < nb_bits; i++) {
316 node.fingers[i].id = -1;
317 set_finger(&node, i, node.id);
320 if (argc == 3) { // first ring
321 deadline = xbt_str_parse_double(argv[2],"Invalid deadline: %s");
326 int known_id = xbt_str_parse_int(argv[2],"Invalid root ID: %s");
327 //double sleep_time = atof(argv[3]);
328 deadline = xbt_str_parse_double(argv[4],"Invalid deadline: %s");
331 // sleep before starting
332 XBT_DEBUG("Let's sleep during %f", sleep_time);
333 MSG_process_sleep(sleep_time);
335 XBT_DEBUG("Hey! Let's join the system.");
337 join_success = join(&node, known_id);
341 while (MSG_get_clock() < init_time + deadline
342 // && MSG_get_clock() < node.last_change_date + 1000
343 && MSG_get_clock() < max_simulation_time) {
345 if (node.comm_receive == NULL) {
346 task_received = NULL;
347 node.comm_receive = MSG_task_irecv(&task_received, node.mailbox);
348 // FIXME: do not make MSG_task_irecv() calls from several functions
351 //XBT_INFO("Node %d is ring member : %d", node.id, is_ring_member(known_id, node.id) != -1);
353 if (!MSG_comm_test(node.comm_receive)) {
355 // no task was received: make some periodic calls
357 if(MC_is_active() || MC_record_replay_is_active()){
358 if(MC_is_active() && !MC_visited_reduction() && no_op){
361 if(listen == 0 && (sub_protocol = MC_random(0, 4)) > 0){
362 if(sub_protocol == 1)
364 else if(sub_protocol == 2)
366 else if(sub_protocol == 3)
367 check_predecessor(&node);
369 random_lookup(&node);
372 MSG_process_sleep(sleep_delay);
373 if(!MC_visited_reduction())
377 if (MSG_get_clock() >= next_stabilize_date) {
379 next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay;
380 }else if (MSG_get_clock() >= next_fix_fingers_date) {
382 next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay;
383 }else if (MSG_get_clock() >= next_check_predecessor_date) {
384 check_predecessor(&node);
385 next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay;
386 }else if (MSG_get_clock() >= next_lookup_date) {
387 random_lookup(&node);
388 next_lookup_date = MSG_get_clock() + periodic_lookup_delay;
390 // nothing to do: sleep for a while
391 MSG_process_sleep(sleep_delay);
396 // a transfer has occurred
398 msg_error_t status = MSG_comm_get_status(node.comm_receive);
400 if (status != MSG_OK) {
401 XBT_DEBUG("Failed to receive a task. Nevermind.");
402 MSG_comm_destroy(node.comm_receive);
403 node.comm_receive = NULL;
406 // the task was successfully received
407 MSG_comm_destroy(node.comm_receive);
408 node.comm_receive = NULL;
409 handle_task(&node, task_received);
414 if (node.comm_receive) {
415 /* handle last task if any */
416 if (MSG_comm_wait(node.comm_receive, 0) == MSG_OK)
417 task_free(task_received);
418 MSG_comm_destroy(node.comm_receive);
419 node.comm_receive = NULL;
426 // stop the simulation
427 xbt_free(node.fingers);
432 * \brief This function is called when the current node receives a task.
433 * \param node the current node
434 * \param task the task to handle (don't touch it then:
435 * it will be destroyed, reused or forwarded)
437 static void handle_task(node_t node, msg_task_t task) {
439 XBT_DEBUG("Handling task %p", task);
440 char mailbox[MAILBOX_NAME_SIZE];
441 task_data_t task_data = (task_data_t) MSG_task_get_data(task);
442 e_task_type_t type = task_data->type;
446 case TASK_FIND_SUCCESSOR:
447 XBT_DEBUG("Receiving a 'Find Successor' request from %s for id %d",
448 task_data->issuer_host_name, task_data->request_id);
449 // is my successor the successor?
450 if (is_in_interval(task_data->request_id, node->id + 1, node->fingers[0].id)) {
451 task_data->type = TASK_FIND_SUCCESSOR_ANSWER;
452 task_data->answer_id = node->fingers[0].id;
453 XBT_DEBUG("Sending back a 'Find Successor Answer' to %s (mailbox %s): the successor of %d is %d",
454 task_data->issuer_host_name,
455 task_data->answer_to,
456 task_data->request_id, task_data->answer_id);
457 MSG_task_dsend(task, task_data->answer_to, task_free);
460 // otherwise, forward the request to the closest preceding finger in my table
461 int closest = closest_preceding_node(node, task_data->request_id);
462 XBT_DEBUG("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d",
463 task_data->request_id, closest);
464 get_mailbox(closest, mailbox);
465 MSG_task_dsend(task, mailbox, task_free);
469 case TASK_GET_PREDECESSOR:
470 XBT_DEBUG("Receiving a 'Get Predecessor' request from %s", task_data->issuer_host_name);
471 task_data->type = TASK_GET_PREDECESSOR_ANSWER;
472 task_data->answer_id = node->pred_id;
473 XBT_DEBUG("Sending back a 'Get Predecessor Answer' to %s via mailbox '%s': my predecessor is %d",
474 task_data->issuer_host_name,
475 task_data->answer_to, task_data->answer_id);
476 MSG_task_dsend(task, task_data->answer_to, task_free);
480 // someone is telling me that he may be my new predecessor
481 XBT_DEBUG("Receiving a 'Notify' request from %s", task_data->issuer_host_name);
482 notify(node, task_data->request_id);
486 case TASK_PREDECESSOR_LEAVING:
487 // my predecessor is about to quit
488 XBT_DEBUG("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name);
489 // modify my predecessor
490 set_predecessor(node, task_data->request_id);
493 >> notify my new predecessor
494 >> send a notify_predecessors !!
498 case TASK_SUCCESSOR_LEAVING:
499 // my successor is about to quit
500 XBT_DEBUG("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name);
501 // modify my successor FIXME : this should be implicit ?
502 set_finger(node, 0, task_data->request_id);
505 >> notify my new successor
506 >> update my table & predecessors table */
509 case TASK_FIND_SUCCESSOR_ANSWER:
510 case TASK_GET_PREDECESSOR_ANSWER:
511 case TASK_PREDECESSOR_ALIVE_ANSWER:
512 XBT_DEBUG("Ignoring unexpected task of type %d (%p)", (int)type, task);
516 case TASK_PREDECESSOR_ALIVE:
517 XBT_DEBUG("Receiving a 'Predecessor Alive' request from %s", task_data->issuer_host_name);
518 task_data->type = TASK_PREDECESSOR_ALIVE_ANSWER;
519 XBT_DEBUG("Sending back a 'Predecessor Alive Answer' to %s (mailbox %s)",
520 task_data->issuer_host_name,
521 task_data->answer_to);
522 MSG_task_dsend(task, task_data->answer_to, task_free);
531 * \brief Initializes the current node as the first one of the system.
532 * \param node the current node
534 static void create(node_t node)
536 XBT_DEBUG("Create a new Chord ring...");
537 set_predecessor(node, -1); // -1 means that I have no predecessor
538 print_finger_table(node);
542 * \brief Makes the current node join the ring, knowing the id of a node
543 * already in the ring
544 * \param node the current node
545 * \param known_id id of a node already in the ring
546 * \return 1 if the join operation succeeded, 0 otherwise
548 static int join(node_t node, int known_id)
550 XBT_INFO("Joining the ring with id %d, knowing node %d", node->id, known_id);
551 set_predecessor(node, -1); // no predecessor (yet)
555 for (i = 0; i < nb_bits; i++) {
556 set_finger(node, i, known_id);
560 int successor_id = remote_find_successor(node, known_id, node->id);
561 if (successor_id == -1) {
562 XBT_INFO("Cannot join the ring.");
565 set_finger(node, 0, successor_id);
566 print_finger_table(node);
569 return successor_id != -1;
573 * \brief Makes the current node quit the system
574 * \param node the current node
576 static void leave(node_t node)
578 XBT_DEBUG("Well Guys! I Think it's time for me to quit ;)");
583 * \brief Notifies the successor and the predecessor of the current node
585 * \param node the current node
587 static void quit_notify(node_t node)
589 char mailbox[MAILBOX_NAME_SIZE];
590 //send the PREDECESSOR_LEAVING to our successor
591 task_data_t req_data = xbt_new0(s_task_data_t,1);
592 req_data->type = TASK_PREDECESSOR_LEAVING;
593 req_data->request_id = node->pred_id;
594 get_mailbox(node->id, req_data->answer_to);
595 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
597 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
598 XBT_DEBUG("Sending a 'PREDECESSOR_LEAVING' to my successor %d",node->fingers[0].id);
599 if (MSG_task_send_with_timeout(task_sent, node->fingers[0].mailbox, timeout)==
601 XBT_DEBUG("Timeout expired when sending a 'PREDECESSOR_LEAVING' to my successor %d",
602 node->fingers[0].id);
603 task_free(task_sent);
606 //send the SUCCESSOR_LEAVING to our predecessor
607 get_mailbox(node->pred_id, mailbox);
608 task_data_t req_data_s = xbt_new0(s_task_data_t,1);
609 req_data_s->type = TASK_SUCCESSOR_LEAVING;
610 req_data_s->request_id = node->fingers[0].id;
611 req_data_s->request_id = node->pred_id;
612 get_mailbox(node->id, req_data_s->answer_to);
613 req_data_s->issuer_host_name = MSG_host_get_name(MSG_host_self());
615 msg_task_t task_sent_s = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data_s);
616 XBT_DEBUG("Sending a 'SUCCESSOR_LEAVING' to my predecessor %d",node->pred_id);
617 if (MSG_task_send_with_timeout(task_sent_s, mailbox, timeout)==
619 XBT_DEBUG("Timeout expired when sending a 'SUCCESSOR_LEAVING' to my predecessor %d",
621 task_free(task_sent_s);
627 * \brief Makes the current node find the successor node of an id.
628 * \param node the current node
629 * \param id the id to find
630 * \return the id of the successor node, or -1 if the request failed
632 static int find_successor(node_t node, int id)
634 // is my successor the successor?
635 if (is_in_interval(id, node->id + 1, node->fingers[0].id)) {
636 return node->fingers[0].id;
639 // otherwise, ask the closest preceding finger in my table
640 int closest = closest_preceding_node(node, id);
641 return remote_find_successor(node, closest, id);
645 * \brief Asks another node the successor node of an id.
646 * \param node the current node
647 * \param ask_to the node to ask to
648 * \param id the id to find
649 * \return the id of the successor node, or -1 if the request failed
651 static int remote_find_successor(node_t node, int ask_to, int id)
655 char mailbox[MAILBOX_NAME_SIZE];
656 get_mailbox(ask_to, mailbox);
657 task_data_t req_data = xbt_new0(s_task_data_t, 1);
658 req_data->type = TASK_FIND_SUCCESSOR;
659 req_data->request_id = id;
660 get_mailbox(node->id, req_data->answer_to);
661 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
663 // send a "Find Successor" request to ask_to_id
664 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
665 XBT_DEBUG("Sending a 'Find Successor' request (task %p) to %d for id %d", task_sent, ask_to, id);
666 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
669 XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
670 task_sent, ask_to, id);
671 task_free(task_sent);
675 // receive the answer
676 XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
677 task_sent, ask_to, id);
680 if (node->comm_receive == NULL) {
681 msg_task_t task_received = NULL;
682 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
685 res = MSG_comm_wait(node->comm_receive, timeout);
688 XBT_DEBUG("Failed to receive the answer to my 'Find Successor' request (task %p): %d",
689 task_sent, (int)res);
691 MSG_comm_destroy(node->comm_receive);
692 node->comm_receive = NULL;
695 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
696 XBT_DEBUG("Received a task (%p)", task_received);
697 task_data_t ans_data = MSG_task_get_data(task_received);
699 // Once upon a time, our code assumed that here, task_received != task_sent all the time
701 // This assumption is wrong (as messages from differing round can interleave), leading to a bug in our code.
702 // We failed to find this bug directly, as it only occured on large platforms, leading to hardly usable traces.
703 // Instead, we used the model-checker to track down the issue by adding the following test here in the code:
704 // if (MC_is_active()) {
705 // MC_assert(task_received == task_sent);
707 // That explained the bug in a snap, with a very cool example and everything.
709 // This MC_assert is now desactivated as the case is now properly handled in our code and we don't want the
710 // MC to fail any further under that condition, but this comment is here to as a memorial for this first
711 // brillant victory of the model-checking in the SimGrid community :)
713 if (task_received != task_sent ||
714 ans_data->type != TASK_FIND_SUCCESSOR_ANSWER) {
715 // this is not the expected answer
716 MSG_comm_destroy(node->comm_receive);
717 node->comm_receive = NULL;
718 handle_task(node, task_received);
721 // this is our answer
722 XBT_DEBUG("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d",
723 ans_data->request_id, task_received, id, ans_data->answer_id);
724 successor = ans_data->answer_id;
726 MSG_comm_destroy(node->comm_receive);
727 node->comm_receive = NULL;
728 task_free(task_received);
738 * \brief Asks another node its predecessor.
739 * \param node the current node
740 * \param ask_to the node to ask to
741 * \return the id of its predecessor node, or -1 if the request failed
742 * (or if the node does not know its predecessor)
744 static int remote_get_predecessor(node_t node, int ask_to)
746 int predecessor_id = -1;
748 char mailbox[MAILBOX_NAME_SIZE];
749 get_mailbox(ask_to, mailbox);
750 task_data_t req_data = xbt_new0(s_task_data_t, 1);
751 req_data->type = TASK_GET_PREDECESSOR;
752 get_mailbox(node->id, req_data->answer_to);
753 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
755 // send a "Get Predecessor" request to ask_to_id
756 XBT_DEBUG("Sending a 'Get Predecessor' request to %d", ask_to);
757 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
758 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
761 XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d",
763 task_free(task_sent);
767 // receive the answer
768 XBT_DEBUG("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
769 task_sent, ask_to, req_data->answer_to);
772 if (node->comm_receive == NULL) { // FIXME simplify this
773 msg_task_t task_received = NULL;
774 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
777 res = MSG_comm_wait(node->comm_receive, timeout);
780 XBT_DEBUG("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d",
781 task_sent, (int)res);
783 MSG_comm_destroy(node->comm_receive);
784 node->comm_receive = NULL;
787 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
788 task_data_t ans_data = MSG_task_get_data(task_received);
790 /*if (MC_is_active()) {
791 MC_assert(task_received == task_sent);
794 if (task_received != task_sent ||
795 ans_data->type != TASK_GET_PREDECESSOR_ANSWER) {
796 MSG_comm_destroy(node->comm_receive);
797 node->comm_receive = NULL;
798 handle_task(node, task_received);
801 XBT_DEBUG("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d",
802 task_received, ask_to, ans_data->answer_id);
803 predecessor_id = ans_data->answer_id;
805 MSG_comm_destroy(node->comm_receive);
806 node->comm_receive = NULL;
807 task_free(task_received);
813 return predecessor_id;
817 * \brief Returns the closest preceding finger of an id
818 * with respect to the finger table of the current node.
819 * \param node the current node
820 * \param id the id to find
821 * \return the closest preceding finger of that id
823 int closest_preceding_node(node_t node, int id)
826 for (i = nb_bits - 1; i >= 0; i--) {
827 if (is_in_interval(node->fingers[i].id, node->id + 1, id - 1)) {
828 return node->fingers[i].id;
835 * \brief This function is called periodically. It checks the immediate
836 * successor of the current node.
837 * \param node the current node
839 static void stabilize(node_t node)
841 XBT_DEBUG("Stabilizing node");
843 // get the predecessor of my immediate successor
845 int successor_id = node->fingers[0].id;
846 if (successor_id != node->id) {
847 candidate_id = remote_get_predecessor(node, successor_id);
850 candidate_id = node->pred_id;
853 // this node is a candidate to become my new successor
854 if (candidate_id != -1
855 && is_in_interval(candidate_id, node->id + 1, successor_id - 1)) {
856 set_finger(node, 0, candidate_id);
858 if (successor_id != node->id) {
859 remote_notify(node, successor_id, node->id);
864 * \brief Notifies the current node that its predecessor may have changed.
865 * \param node the current node
866 * \param candidate_id the possible new predecessor
868 static void notify(node_t node, int predecessor_candidate_id) {
870 if (node->pred_id == -1
871 || is_in_interval(predecessor_candidate_id, node->pred_id + 1, node->id - 1)) {
873 set_predecessor(node, predecessor_candidate_id);
874 print_finger_table(node);
877 XBT_DEBUG("I don't have to change my predecessor to %d", predecessor_candidate_id);
882 * \brief Notifies a remote node that its predecessor may have changed.
883 * \param node the current node
884 * \param notify_id id of the node to notify
885 * \param candidate_id the possible new predecessor
887 static void remote_notify(node_t node, int notify_id, int predecessor_candidate_id) {
889 task_data_t req_data = xbt_new0(s_task_data_t, 1);
890 req_data->type = TASK_NOTIFY;
891 req_data->request_id = predecessor_candidate_id;
892 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
894 // send a "Notify" request to notify_id
895 msg_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
896 XBT_DEBUG("Sending a 'Notify' request (task %p) to %d", task, notify_id);
897 char mailbox[MAILBOX_NAME_SIZE];
898 get_mailbox(notify_id, mailbox);
899 MSG_task_dsend(task, mailbox, task_free);
903 * \brief This function is called periodically.
904 * It refreshes the finger table of the current node.
905 * \param node the current node
907 static void fix_fingers(node_t node) {
909 XBT_DEBUG("Fixing fingers");
910 int i = node->next_finger_to_fix;
911 int id = find_successor(node, node->id + powers2[i]);
914 if (id != node->fingers[i].id) {
915 set_finger(node, i, id);
916 print_finger_table(node);
918 node->next_finger_to_fix = (i + 1) % nb_bits;
923 * \brief This function is called periodically.
924 * It checks whether the predecessor has failed
925 * \param node the current node
927 static void check_predecessor(node_t node)
929 XBT_DEBUG("Checking whether my predecessor is alive");
931 if(node->pred_id == -1)
936 char mailbox[MAILBOX_NAME_SIZE];
937 get_mailbox(node->pred_id, mailbox);
938 task_data_t req_data = xbt_new0(s_task_data_t,1);
939 req_data->type = TASK_PREDECESSOR_ALIVE;
940 req_data->request_id = node->pred_id;
941 get_mailbox(node->id, req_data->answer_to);
942 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
944 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
945 XBT_DEBUG("Sending a 'Predecessor Alive' request to my predecessor %d", node->pred_id);
947 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
950 XBT_DEBUG("Failed to send the 'Predecessor Alive' request (task %p) to %d", task_sent, node->pred_id);
951 task_free(task_sent);
954 // receive the answer
955 XBT_DEBUG("Sent 'Predecessor Alive' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
956 task_sent, node->pred_id, req_data->answer_to);
959 if (node->comm_receive == NULL) { // FIXME simplify this
960 msg_task_t task_received = NULL;
961 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
964 res = MSG_comm_wait(node->comm_receive, timeout);
967 XBT_DEBUG("Failed to receive the answer to my 'Predecessor Alive' request (task %p): %d",
968 task_sent, (int)res);
970 MSG_comm_destroy(node->comm_receive);
971 node->comm_receive = NULL;
974 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
975 if (task_received != task_sent) {
976 MSG_comm_destroy(node->comm_receive);
977 node->comm_receive = NULL;
978 handle_task(node, task_received);
980 XBT_DEBUG("Received the answer to my 'Predecessor Alive' request (task %p) : my predecessor %d is alive",
981 task_received, node->pred_id);
983 MSG_comm_destroy(node->comm_receive);
984 node->comm_receive = NULL;
985 task_free(task_received);
993 * \brief Performs a find successor request to a random id.
994 * \param node the current node
996 static void random_lookup(node_t node)
998 int random_index = RngStream_RandInt (node->stream, 0, nb_bits - 1);
999 int random_id = node->fingers[random_index].id;
1000 XBT_DEBUG("Making a lookup request for id %d", random_id);
1001 int res = find_successor(node, random_id);
1002 XBT_DEBUG("The successor of node %d is %d", random_id, res);
1006 int main(int argc, char *argv[])
1008 MSG_init(&argc, argv);
1009 xbt_assert(argc > 2,
1010 "Usage: %s [-nb_bits=n] [-timeout=t] platform_file deployment_file\n"
1011 "\tExample: %s ../msg_platform.xml chord.xml\n", argv[0], argv[0]);
1013 char **options = &argv[1];
1014 while (!strncmp(options[0], "-", 1)) {
1016 int length = strlen("-nb_bits=");
1017 if (!strncmp(options[0], "-nb_bits=", length) && strlen(options[0]) > length) {
1018 nb_bits = xbt_str_parse_int(options[0] + length, "Invalid nb_bits parameter: %s");
1019 XBT_DEBUG("Set nb_bits to %d", nb_bits);
1023 length = strlen("-timeout=");
1024 if (!strncmp(options[0], "-timeout=", length) && strlen(options[0]) > length) {
1025 timeout = xbt_str_parse_int(options[0] + length, "Invalid timeout parameter: %s");
1026 XBT_DEBUG("Set timeout to %d", timeout);
1029 xbt_die("Invalid chord option '%s'", options[0]);
1035 const char* platform_file = options[0];
1036 const char* application_file = options[1];
1038 MSG_create_environment(platform_file);
1042 MSG_function_register("node", node);
1043 MSG_launch_application(application_file);
1045 msg_error_t res = MSG_main();
1046 XBT_CRITICAL("Messages created: %ld", smx_total_comms);
1047 XBT_INFO("Simulated time: %g", MSG_get_clock());
1051 return res != MSG_OK;