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;
38 typedef struct s_finger {
40 char mailbox[MAILBOX_NAME_SIZE]; // string representation of the id
41 } s_finger_t, *finger_t;
44 typedef struct s_node {
46 char mailbox[MAILBOX_NAME_SIZE]; // my mailbox name (string representation of the id)
47 s_finger_t *fingers; // finger table, of size nb_bits (fingers[0] is my successor)
48 int pred_id; // predecessor id
49 char pred_mailbox[MAILBOX_NAME_SIZE]; // predecessor's mailbox name
50 int next_finger_to_fix; // index of the next finger to fix in fix_fingers()
51 msg_comm_t comm_receive; // current communication to receive
52 double last_change_date; // last time I changed a finger or my predecessor
53 RngStream stream; //RngStream for
56 /* Types of tasks exchanged between nodes. */
59 TASK_FIND_SUCCESSOR_ANSWER,
61 TASK_GET_PREDECESSOR_ANSWER,
63 TASK_SUCCESSOR_LEAVING,
64 TASK_PREDECESSOR_LEAVING,
65 TASK_PREDECESSOR_ALIVE,
66 TASK_PREDECESSOR_ALIVE_ANSWER
69 /* Data attached with the tasks sent and received */
70 typedef struct s_task_data {
71 e_task_type_t type; // type of task
72 int request_id; // id paramater (used by some types of tasks)
73 int request_finger; // finger parameter (used by some types of tasks)
74 int answer_id; // answer (used by some types of tasks)
75 char answer_to[MAILBOX_NAME_SIZE]; // mailbox to send an answer to (if any)
76 const char* issuer_host_name; // used for logging
77 } s_task_data_t, *task_data_t;
80 static xbt_dynar_t host_list;
83 static void chord_exit(void);
84 static int normalize(int id);
85 static int is_in_interval(int id, int start, int end);
86 static void get_mailbox(int host_id, char* mailbox);
87 static void task_free(void* task);
88 static void print_finger_table(node_t node);
89 static void set_finger(node_t node, int finger_index, int id);
90 static void set_predecessor(node_t node, int predecessor_id);
93 static int node(int argc, char *argv[]);
94 static void handle_task(node_t node, msg_task_t task);
97 static void create(node_t node);
98 static int join(node_t node, int known_id);
99 static void leave(node_t node);
100 static int find_successor(node_t node, int id);
101 static int remote_find_successor(node_t node, int ask_to_id, int id);
102 static int remote_get_predecessor(node_t node, int ask_to_id);
103 static int closest_preceding_node(node_t node, int id);
104 static void stabilize(node_t node);
105 static void notify(node_t node, int predecessor_candidate_id);
106 static void remote_notify(node_t node, int notify_to, int predecessor_candidate_id);
107 static void fix_fingers(node_t node);
108 static void check_predecessor(node_t node);
109 static void random_lookup(node_t);
110 static void quit_notify(node_t node);
112 /* \brief Global initialization of the Chord simulation. */
113 static void chord_initialize(void)
115 // compute the powers of 2 once for all
116 powers2 = xbt_new(int, nb_bits);
119 for (i = 0; i < nb_bits; i++) {
124 XBT_DEBUG("Sets nb_keys to %d", nb_keys);
127 host_list = MSG_hosts_as_dynar();
128 xbt_dynar_foreach(host_list, i, host) {
131 snprintf(descr, sizeof descr, "RngSream<%s>", MSG_host_get_name(host));
132 stream = RngStream_CreateStream(descr);
133 MSG_host_set_property_value(host, "stream", (char*)stream, NULL);
137 static void chord_exit(void)
141 xbt_dynar_foreach(host_list, i, host) {
142 RngStream stream = (RngStream)MSG_host_get_property_value(host, "stream");
143 RngStream_DeleteStream(&stream);
145 xbt_dynar_free(&host_list);
151 * \brief Turns an id into an equivalent id in [0, nb_keys).
153 * \return the corresponding normalized id
155 static int normalize(int id)
157 // like id % nb_keys, but works with negatives numbers (and faster)
158 return id & (nb_keys - 1);
162 * \brief Returns whether an id belongs to the interval [start, end].
164 * The parameters are noramlized to make sure they are between 0 and nb_keys - 1).
165 * 1 belongs to [62, 3]
166 * 1 does not belong to [3, 62]
167 * 63 belongs to [62, 3]
168 * 63 does not belong to [3, 62]
169 * 24 belongs to [21, 29]
170 * 24 does not belong to [29, 21]
172 * \param id id to check
173 * \param start lower bound
174 * \param end upper bound
175 * \return a non-zero value if id in in [start, end]
177 static int is_in_interval(int id, int start, int end)
180 start = normalize(start);
181 end = normalize(end);
183 // make sure end >= start and id >= start
196 * \brief Gets the mailbox name of a host given its chord id.
197 * \param node_id id of a node
198 * \param mailbox pointer to where the mailbox name should be written
199 * (there must be enough space)
201 static void get_mailbox(int node_id, char* mailbox)
203 snprintf(mailbox, MAILBOX_NAME_SIZE - 1, "%d", node_id);
207 * \brief Frees the memory used by a task.
208 * \param task the MSG task to destroy
210 static void task_free(void* task)
212 // TODO add a parameter data_free_function to MSG_task_create?
214 xbt_free(MSG_task_get_data(task));
215 MSG_task_destroy(task);
220 * \brief Displays the finger table of a node.
223 static void print_finger_table(node_t node)
225 if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) {
227 XBT_VERB("My finger table:");
228 XBT_VERB("Start | Succ");
229 for (i = 0; i < nb_bits; i++) {
230 XBT_VERB(" %3d | %3d", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
232 XBT_VERB("Predecessor: %d", node->pred_id);
237 * \brief Sets a finger of the current node.
238 * \param node the current node
239 * \param finger_index index of the finger to set (0 to nb_bits - 1)
240 * \param id the id to set for this finger
242 static void set_finger(node_t node, int finger_index, int id)
244 if (id != node->fingers[finger_index].id) {
245 node->fingers[finger_index].id = id;
246 get_mailbox(id, node->fingers[finger_index].mailbox);
247 node->last_change_date = MSG_get_clock();
248 XBT_DEBUG("My new finger #%d is %d", finger_index, id);
253 * \brief Sets the predecessor of the current node.
254 * \param node the current node
255 * \param id the id to predecessor, or -1 to unset the predecessor
257 static void set_predecessor(node_t node, int predecessor_id)
259 if (predecessor_id != node->pred_id) {
260 node->pred_id = predecessor_id;
262 if (predecessor_id != -1) {
263 get_mailbox(predecessor_id, node->pred_mailbox);
265 node->last_change_date = MSG_get_clock();
267 XBT_DEBUG("My new predecessor is %d", predecessor_id);
272 * \brief Node Function
275 * - the id of a guy I know in the system (except for the first node)
276 * - the time to sleep before I join (except for the first node)
278 int node(int argc, char *argv[])
281 /* Reduce the run size for the MC */
282 if(MC_is_active() || MC_record_replay_is_active()){
283 periodic_stabilize_delay = 8;
284 periodic_fix_fingers_delay = 8;
285 periodic_check_predecessor_delay = 8;
288 double init_time = MSG_get_clock();
289 msg_task_t task_received = NULL;
291 int join_success = 0;
293 double next_stabilize_date = init_time + periodic_stabilize_delay;
294 double next_fix_fingers_date = init_time + periodic_fix_fingers_delay;
295 double next_check_predecessor_date = init_time + periodic_check_predecessor_delay;
296 double next_lookup_date = init_time + periodic_lookup_delay;
300 int sub_protocol = 0;
302 xbt_assert(argc == 3 || argc == 5, "Wrong number of arguments for this node");
304 // initialize my node
306 node.id = xbt_str_parse_int(argv[1],"Invalid ID: %s");
307 node.stream = (RngStream)MSG_host_get_property_value(MSG_host_self(), "stream");
308 get_mailbox(node.id, node.mailbox);
309 node.next_finger_to_fix = 0;
310 node.fingers = xbt_new0(s_finger_t, nb_bits);
311 node.last_change_date = init_time;
313 for (i = 0; i < nb_bits; i++) {
314 node.fingers[i].id = -1;
315 set_finger(&node, i, node.id);
318 if (argc == 3) { // first ring
319 deadline = xbt_str_parse_double(argv[2],"Invalid deadline: %s");
324 int known_id = xbt_str_parse_int(argv[2],"Invalid root ID: %s");
325 //double sleep_time = atof(argv[3]);
326 deadline = xbt_str_parse_double(argv[4],"Invalid deadline: %s");
329 // sleep before starting
330 XBT_DEBUG("Let's sleep during %f", sleep_time);
331 MSG_process_sleep(sleep_time);
333 XBT_DEBUG("Hey! Let's join the system.");
335 join_success = join(&node, known_id);
339 while (MSG_get_clock() < init_time + deadline
340 // && MSG_get_clock() < node.last_change_date + 1000
341 && MSG_get_clock() < max_simulation_time) {
343 if (node.comm_receive == NULL) {
344 task_received = NULL;
345 node.comm_receive = MSG_task_irecv(&task_received, node.mailbox);
346 // FIXME: do not make MSG_task_irecv() calls from several functions
349 //XBT_INFO("Node %d is ring member : %d", node.id, is_ring_member(known_id, node.id) != -1);
351 if (!MSG_comm_test(node.comm_receive)) {
353 // no task was received: make some periodic calls
355 if(MC_is_active() || MC_record_replay_is_active()){
356 if(MC_is_active() && !MC_visited_reduction() && no_op){
359 if(listen == 0 && (sub_protocol = MC_random(0, 4)) > 0){
360 if(sub_protocol == 1)
362 else if(sub_protocol == 2)
364 else if(sub_protocol == 3)
365 check_predecessor(&node);
367 random_lookup(&node);
370 MSG_process_sleep(sleep_delay);
371 if(!MC_visited_reduction())
375 if (MSG_get_clock() >= next_stabilize_date) {
377 next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay;
378 }else if (MSG_get_clock() >= next_fix_fingers_date) {
380 next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay;
381 }else if (MSG_get_clock() >= next_check_predecessor_date) {
382 check_predecessor(&node);
383 next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay;
384 }else if (MSG_get_clock() >= next_lookup_date) {
385 random_lookup(&node);
386 next_lookup_date = MSG_get_clock() + periodic_lookup_delay;
388 // nothing to do: sleep for a while
389 MSG_process_sleep(sleep_delay);
394 // a transfer has occurred
396 msg_error_t status = MSG_comm_get_status(node.comm_receive);
398 if (status != MSG_OK) {
399 XBT_DEBUG("Failed to receive a task. Nevermind.");
400 MSG_comm_destroy(node.comm_receive);
401 node.comm_receive = NULL;
404 // the task was successfully received
405 MSG_comm_destroy(node.comm_receive);
406 node.comm_receive = NULL;
407 handle_task(&node, task_received);
412 if (node.comm_receive) {
413 /* handle last task if any */
414 if (MSG_comm_wait(node.comm_receive, 0) == MSG_OK)
415 task_free(task_received);
416 MSG_comm_destroy(node.comm_receive);
417 node.comm_receive = NULL;
424 // stop the simulation
425 xbt_free(node.fingers);
430 * \brief This function is called when the current node receives a task.
431 * \param node the current node
432 * \param task the task to handle (don't touch it then:
433 * it will be destroyed, reused or forwarded)
435 static void handle_task(node_t node, msg_task_t task) {
437 XBT_DEBUG("Handling task %p", task);
438 char mailbox[MAILBOX_NAME_SIZE];
439 task_data_t task_data = (task_data_t) MSG_task_get_data(task);
440 e_task_type_t type = task_data->type;
444 case TASK_FIND_SUCCESSOR:
445 XBT_DEBUG("Receiving a 'Find Successor' request from %s for id %d",
446 task_data->issuer_host_name, task_data->request_id);
447 // is my successor the successor?
448 if (is_in_interval(task_data->request_id, node->id + 1, node->fingers[0].id)) {
449 task_data->type = TASK_FIND_SUCCESSOR_ANSWER;
450 task_data->answer_id = node->fingers[0].id;
451 XBT_DEBUG("Sending back a 'Find Successor Answer' to %s (mailbox %s): the successor of %d is %d",
452 task_data->issuer_host_name,
453 task_data->answer_to,
454 task_data->request_id, task_data->answer_id);
455 MSG_task_dsend(task, task_data->answer_to, task_free);
458 // otherwise, forward the request to the closest preceding finger in my table
459 int closest = closest_preceding_node(node, task_data->request_id);
460 XBT_DEBUG("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d",
461 task_data->request_id, closest);
462 get_mailbox(closest, mailbox);
463 MSG_task_dsend(task, mailbox, task_free);
467 case TASK_GET_PREDECESSOR:
468 XBT_DEBUG("Receiving a 'Get Predecessor' request from %s", task_data->issuer_host_name);
469 task_data->type = TASK_GET_PREDECESSOR_ANSWER;
470 task_data->answer_id = node->pred_id;
471 XBT_DEBUG("Sending back a 'Get Predecessor Answer' to %s via mailbox '%s': my predecessor is %d",
472 task_data->issuer_host_name,
473 task_data->answer_to, task_data->answer_id);
474 MSG_task_dsend(task, task_data->answer_to, task_free);
478 // someone is telling me that he may be my new predecessor
479 XBT_DEBUG("Receiving a 'Notify' request from %s", task_data->issuer_host_name);
480 notify(node, task_data->request_id);
484 case TASK_PREDECESSOR_LEAVING:
485 // my predecessor is about to quit
486 XBT_DEBUG("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name);
487 // modify my predecessor
488 set_predecessor(node, task_data->request_id);
491 >> notify my new predecessor
492 >> send a notify_predecessors !!
496 case TASK_SUCCESSOR_LEAVING:
497 // my successor is about to quit
498 XBT_DEBUG("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name);
499 // modify my successor FIXME : this should be implicit ?
500 set_finger(node, 0, task_data->request_id);
503 >> notify my new successor
504 >> update my table & predecessors table */
507 case TASK_FIND_SUCCESSOR_ANSWER:
508 case TASK_GET_PREDECESSOR_ANSWER:
509 case TASK_PREDECESSOR_ALIVE_ANSWER:
510 XBT_DEBUG("Ignoring unexpected task of type %d (%p)", (int)type, task);
514 case TASK_PREDECESSOR_ALIVE:
515 XBT_DEBUG("Receiving a 'Predecessor Alive' request from %s", task_data->issuer_host_name);
516 task_data->type = TASK_PREDECESSOR_ALIVE_ANSWER;
517 XBT_DEBUG("Sending back a 'Predecessor Alive Answer' to %s (mailbox %s)",
518 task_data->issuer_host_name,
519 task_data->answer_to);
520 MSG_task_dsend(task, task_data->answer_to, task_free);
529 * \brief Initializes the current node as the first one of the system.
530 * \param node the current node
532 static void create(node_t node)
534 XBT_DEBUG("Create a new Chord ring...");
535 set_predecessor(node, -1); // -1 means that I have no predecessor
536 print_finger_table(node);
540 * \brief Makes the current node join the ring, knowing the id of a node
541 * already in the ring
542 * \param node the current node
543 * \param known_id id of a node already in the ring
544 * \return 1 if the join operation succeeded, 0 otherwise
546 static int join(node_t node, int known_id)
548 XBT_INFO("Joining the ring with id %d, knowing node %d", node->id, known_id);
549 set_predecessor(node, -1); // no predecessor (yet)
553 for (i = 0; i < nb_bits; i++) {
554 set_finger(node, i, known_id);
558 int successor_id = remote_find_successor(node, known_id, node->id);
559 if (successor_id == -1) {
560 XBT_INFO("Cannot join the ring.");
563 set_finger(node, 0, successor_id);
564 print_finger_table(node);
567 return successor_id != -1;
571 * \brief Makes the current node quit the system
572 * \param node the current node
574 static void leave(node_t node)
576 XBT_DEBUG("Well Guys! I Think it's time for me to quit ;)");
581 * \brief Notifies the successor and the predecessor of the current node
583 * \param node the current node
585 static void quit_notify(node_t node)
587 char mailbox[MAILBOX_NAME_SIZE];
588 //send the PREDECESSOR_LEAVING to our successor
589 task_data_t req_data = xbt_new0(s_task_data_t,1);
590 req_data->type = TASK_PREDECESSOR_LEAVING;
591 req_data->request_id = node->pred_id;
592 get_mailbox(node->id, req_data->answer_to);
593 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
595 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
596 XBT_DEBUG("Sending a 'PREDECESSOR_LEAVING' to my successor %d",node->fingers[0].id);
597 if (MSG_task_send_with_timeout(task_sent, node->fingers[0].mailbox, timeout)==
599 XBT_DEBUG("Timeout expired when sending a 'PREDECESSOR_LEAVING' to my successor %d",
600 node->fingers[0].id);
601 task_free(task_sent);
604 //send the SUCCESSOR_LEAVING to our predecessor
605 get_mailbox(node->pred_id, mailbox);
606 task_data_t req_data_s = xbt_new0(s_task_data_t,1);
607 req_data_s->type = TASK_SUCCESSOR_LEAVING;
608 req_data_s->request_id = node->fingers[0].id;
609 req_data_s->request_id = node->pred_id;
610 get_mailbox(node->id, req_data_s->answer_to);
611 req_data_s->issuer_host_name = MSG_host_get_name(MSG_host_self());
613 msg_task_t task_sent_s = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data_s);
614 XBT_DEBUG("Sending a 'SUCCESSOR_LEAVING' to my predecessor %d",node->pred_id);
615 if (MSG_task_send_with_timeout(task_sent_s, mailbox, timeout)==
617 XBT_DEBUG("Timeout expired when sending a 'SUCCESSOR_LEAVING' to my predecessor %d",
619 task_free(task_sent_s);
625 * \brief Makes the current node find the successor node of an id.
626 * \param node the current node
627 * \param id the id to find
628 * \return the id of the successor node, or -1 if the request failed
630 static int find_successor(node_t node, int id)
632 // is my successor the successor?
633 if (is_in_interval(id, node->id + 1, node->fingers[0].id)) {
634 return node->fingers[0].id;
637 // otherwise, ask the closest preceding finger in my table
638 int closest = closest_preceding_node(node, id);
639 return remote_find_successor(node, closest, id);
643 * \brief Asks another node the successor node of an id.
644 * \param node the current node
645 * \param ask_to the node to ask to
646 * \param id the id to find
647 * \return the id of the successor node, or -1 if the request failed
649 static int remote_find_successor(node_t node, int ask_to, int id)
653 char mailbox[MAILBOX_NAME_SIZE];
654 get_mailbox(ask_to, mailbox);
655 task_data_t req_data = xbt_new0(s_task_data_t, 1);
656 req_data->type = TASK_FIND_SUCCESSOR;
657 req_data->request_id = id;
658 get_mailbox(node->id, req_data->answer_to);
659 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
661 // send a "Find Successor" request to ask_to_id
662 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
663 XBT_DEBUG("Sending a 'Find Successor' request (task %p) to %d for id %d", task_sent, ask_to, id);
664 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
667 XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
668 task_sent, ask_to, id);
669 task_free(task_sent);
673 // receive the answer
674 XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
675 task_sent, ask_to, id);
678 if (node->comm_receive == NULL) {
679 msg_task_t task_received = NULL;
680 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
683 res = MSG_comm_wait(node->comm_receive, timeout);
686 XBT_DEBUG("Failed to receive the answer to my 'Find Successor' request (task %p): %d",
687 task_sent, (int)res);
689 MSG_comm_destroy(node->comm_receive);
690 node->comm_receive = NULL;
693 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
694 XBT_DEBUG("Received a task (%p)", task_received);
695 task_data_t ans_data = MSG_task_get_data(task_received);
697 // Once upon a time, our code assumed that here, task_received != task_sent all the time
699 // This assumption is wrong (as messages from differing round can interleave), leading to a bug in our code.
700 // We failed to find this bug directly, as it only occured on large platforms, leading to hardly usable traces.
701 // Instead, we used the model-checker to track down the issue by adding the following test here in the code:
702 // if (MC_is_active()) {
703 // MC_assert(task_received == task_sent);
705 // That explained the bug in a snap, with a very cool example and everything.
707 // This MC_assert is now desactivated as the case is now properly handled in our code and we don't want the
708 // MC to fail any further under that condition, but this comment is here to as a memorial for this first
709 // brillant victory of the model-checking in the SimGrid community :)
711 if (task_received != task_sent ||
712 ans_data->type != TASK_FIND_SUCCESSOR_ANSWER) {
713 // this is not the expected answer
714 MSG_comm_destroy(node->comm_receive);
715 node->comm_receive = NULL;
716 handle_task(node, task_received);
719 // this is our answer
720 XBT_DEBUG("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d",
721 ans_data->request_id, task_received, id, ans_data->answer_id);
722 successor = ans_data->answer_id;
724 MSG_comm_destroy(node->comm_receive);
725 node->comm_receive = NULL;
726 task_free(task_received);
736 * \brief Asks another node its predecessor.
737 * \param node the current node
738 * \param ask_to the node to ask to
739 * \return the id of its predecessor node, or -1 if the request failed
740 * (or if the node does not know its predecessor)
742 static int remote_get_predecessor(node_t node, int ask_to)
744 int predecessor_id = -1;
746 char mailbox[MAILBOX_NAME_SIZE];
747 get_mailbox(ask_to, mailbox);
748 task_data_t req_data = xbt_new0(s_task_data_t, 1);
749 req_data->type = TASK_GET_PREDECESSOR;
750 get_mailbox(node->id, req_data->answer_to);
751 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
753 // send a "Get Predecessor" request to ask_to_id
754 XBT_DEBUG("Sending a 'Get Predecessor' request to %d", ask_to);
755 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
756 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
759 XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d",
761 task_free(task_sent);
765 // receive the answer
766 XBT_DEBUG("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
767 task_sent, ask_to, req_data->answer_to);
770 if (node->comm_receive == NULL) { // FIXME simplify this
771 msg_task_t task_received = NULL;
772 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
775 res = MSG_comm_wait(node->comm_receive, timeout);
778 XBT_DEBUG("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d",
779 task_sent, (int)res);
781 MSG_comm_destroy(node->comm_receive);
782 node->comm_receive = NULL;
785 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
786 task_data_t ans_data = MSG_task_get_data(task_received);
788 /*if (MC_is_active()) {
789 MC_assert(task_received == task_sent);
792 if (task_received != task_sent ||
793 ans_data->type != TASK_GET_PREDECESSOR_ANSWER) {
794 MSG_comm_destroy(node->comm_receive);
795 node->comm_receive = NULL;
796 handle_task(node, task_received);
799 XBT_DEBUG("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d",
800 task_received, ask_to, ans_data->answer_id);
801 predecessor_id = ans_data->answer_id;
803 MSG_comm_destroy(node->comm_receive);
804 node->comm_receive = NULL;
805 task_free(task_received);
811 return predecessor_id;
815 * \brief Returns the closest preceding finger of an id
816 * with respect to the finger table of the current node.
817 * \param node the current node
818 * \param id the id to find
819 * \return the closest preceding finger of that id
821 int closest_preceding_node(node_t node, int id)
824 for (i = nb_bits - 1; i >= 0; i--) {
825 if (is_in_interval(node->fingers[i].id, node->id + 1, id - 1)) {
826 return node->fingers[i].id;
833 * \brief This function is called periodically. It checks the immediate
834 * successor of the current node.
835 * \param node the current node
837 static void stabilize(node_t node)
839 XBT_DEBUG("Stabilizing node");
841 // get the predecessor of my immediate successor
843 int successor_id = node->fingers[0].id;
844 if (successor_id != node->id) {
845 candidate_id = remote_get_predecessor(node, successor_id);
848 candidate_id = node->pred_id;
851 // this node is a candidate to become my new successor
852 if (candidate_id != -1
853 && is_in_interval(candidate_id, node->id + 1, successor_id - 1)) {
854 set_finger(node, 0, candidate_id);
856 if (successor_id != node->id) {
857 remote_notify(node, successor_id, node->id);
862 * \brief Notifies the current node that its predecessor may have changed.
863 * \param node the current node
864 * \param candidate_id the possible new predecessor
866 static void notify(node_t node, int predecessor_candidate_id) {
868 if (node->pred_id == -1
869 || is_in_interval(predecessor_candidate_id, node->pred_id + 1, node->id - 1)) {
871 set_predecessor(node, predecessor_candidate_id);
872 print_finger_table(node);
875 XBT_DEBUG("I don't have to change my predecessor to %d", predecessor_candidate_id);
880 * \brief Notifies a remote node that its predecessor may have changed.
881 * \param node the current node
882 * \param notify_id id of the node to notify
883 * \param candidate_id the possible new predecessor
885 static void remote_notify(node_t node, int notify_id, int predecessor_candidate_id) {
887 task_data_t req_data = xbt_new0(s_task_data_t, 1);
888 req_data->type = TASK_NOTIFY;
889 req_data->request_id = predecessor_candidate_id;
890 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
892 // send a "Notify" request to notify_id
893 msg_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
894 XBT_DEBUG("Sending a 'Notify' request (task %p) to %d", task, notify_id);
895 char mailbox[MAILBOX_NAME_SIZE];
896 get_mailbox(notify_id, mailbox);
897 MSG_task_dsend(task, mailbox, task_free);
901 * \brief This function is called periodically.
902 * It refreshes the finger table of the current node.
903 * \param node the current node
905 static void fix_fingers(node_t node) {
907 XBT_DEBUG("Fixing fingers");
908 int i = node->next_finger_to_fix;
909 int id = find_successor(node, node->id + powers2[i]);
912 if (id != node->fingers[i].id) {
913 set_finger(node, i, id);
914 print_finger_table(node);
916 node->next_finger_to_fix = (i + 1) % nb_bits;
921 * \brief This function is called periodically.
922 * It checks whether the predecessor has failed
923 * \param node the current node
925 static void check_predecessor(node_t node)
927 XBT_DEBUG("Checking whether my predecessor is alive");
929 if(node->pred_id == -1)
934 char mailbox[MAILBOX_NAME_SIZE];
935 get_mailbox(node->pred_id, mailbox);
936 task_data_t req_data = xbt_new0(s_task_data_t,1);
937 req_data->type = TASK_PREDECESSOR_ALIVE;
938 req_data->request_id = node->pred_id;
939 get_mailbox(node->id, req_data->answer_to);
940 req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
942 msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
943 XBT_DEBUG("Sending a 'Predecessor Alive' request to my predecessor %d", node->pred_id);
945 msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout);
948 XBT_DEBUG("Failed to send the 'Predecessor Alive' request (task %p) to %d", task_sent, node->pred_id);
949 task_free(task_sent);
952 // receive the answer
953 XBT_DEBUG("Sent 'Predecessor Alive' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
954 task_sent, node->pred_id, req_data->answer_to);
957 if (node->comm_receive == NULL) { // FIXME simplify this
958 msg_task_t task_received = NULL;
959 node->comm_receive = MSG_task_irecv(&task_received, node->mailbox);
962 res = MSG_comm_wait(node->comm_receive, timeout);
965 XBT_DEBUG("Failed to receive the answer to my 'Predecessor Alive' request (task %p): %d",
966 task_sent, (int)res);
968 MSG_comm_destroy(node->comm_receive);
969 node->comm_receive = NULL;
972 msg_task_t task_received = MSG_comm_get_task(node->comm_receive);
973 if (task_received != task_sent) {
974 MSG_comm_destroy(node->comm_receive);
975 node->comm_receive = NULL;
976 handle_task(node, task_received);
978 XBT_DEBUG("Received the answer to my 'Predecessor Alive' request (task %p) : my predecessor %d is alive",
979 task_received, node->pred_id);
981 MSG_comm_destroy(node->comm_receive);
982 node->comm_receive = NULL;
983 task_free(task_received);
991 * \brief Performs a find successor request to a random id.
992 * \param node the current node
994 static void random_lookup(node_t node)
996 int random_index = RngStream_RandInt (node->stream, 0, nb_bits - 1);
997 int random_id = node->fingers[random_index].id;
998 XBT_DEBUG("Making a lookup request for id %d", random_id);
999 int res = find_successor(node, random_id);
1000 XBT_DEBUG("The successor of node %d is %d", random_id, res);
1004 int main(int argc, char *argv[])
1006 MSG_init(&argc, argv);
1007 xbt_assert(argc > 2,
1008 "Usage: %s [-nb_bits=n] [-timeout=t] platform_file deployment_file\n"
1009 "\tExample: %s ../msg_platform.xml chord.xml\n", argv[0], argv[0]);
1011 char **options = &argv[1];
1012 while (!strncmp(options[0], "-", 1)) {
1014 int length = strlen("-nb_bits=");
1015 if (!strncmp(options[0], "-nb_bits=", length) && strlen(options[0]) > length) {
1016 nb_bits = xbt_str_parse_int(options[0] + length, "Invalid nb_bits parameter: %s");
1017 XBT_DEBUG("Set nb_bits to %d", nb_bits);
1021 length = strlen("-timeout=");
1022 if (!strncmp(options[0], "-timeout=", length) && strlen(options[0]) > length) {
1023 timeout = xbt_str_parse_int(options[0] + length, "Invalid timeout parameter: %s");
1024 XBT_DEBUG("Set timeout to %d", timeout);
1027 xbt_die("Invalid chord option '%s'", options[0]);
1033 const char* platform_file = options[0];
1034 const char* application_file = options[1];
1036 MSG_create_environment(platform_file);
1040 MSG_function_register("node", node);
1041 MSG_launch_application(application_file);
1043 msg_error_t res = MSG_main();
1044 XBT_INFO("Simulated time: %g", MSG_get_clock());
1048 return res != MSG_OK;