X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/7b28e9e1bfe2348d1a820e78d8867ba325dcd03b..8feb1c4d2fe696b8f0c723eac3d88024976f6825:/examples/msg/chord/chord.c diff --git a/examples/msg/chord/chord.c b/examples/msg/chord/chord.c index 7b9bf0eed9..7289d43b36 100644 --- a/examples/msg/chord/chord.c +++ b/examples/msg/chord/chord.c @@ -1,5 +1,5 @@ -/* Copyright (c) 2010. The SimGrid Team. +/* Copyright (c) 2010-2013. The SimGrid Team. * All rights reserved. */ /* This program is free software; you can redistribute it and/or modify it @@ -9,64 +9,107 @@ #include "msg/msg.h" #include "xbt/log.h" #include "xbt/asserts.h" +#include "simgrid/modelchecker.h" +#include + +/** @addtogroup MSG_examples + * + * - chord/chord.c: Classical Chord P2P protocol + * This example implements the well known Chord P2P protocol. Its + * main advantage is that it constitute a fully working non-trivial + * example. In addition, its implementation is rather efficient, as + * demonstrated in http://hal.inria.fr/inria-00602216/ + */ + + XBT_LOG_NEW_DEFAULT_CATEGORY(msg_chord, "Messages specific for this msg example"); -#define NB_BITS 6 -#define NB_KEYS 64 +#define COMM_SIZE 10 +#define COMP_SIZE 0 +#define MAILBOX_NAME_SIZE 10 -/** +static int nb_bits = 24; +static int nb_keys = 0; +static int timeout = 50; +static int max_simulation_time = 1000; +static int periodic_stabilize_delay = 20; +static int periodic_fix_fingers_delay = 120; +static int periodic_check_predecessor_delay = 120; +static int periodic_lookup_delay = 10; + +extern long int smx_total_comms; + +/* * Finger element. */ -typedef struct finger { +typedef struct s_finger { int id; - char* mailbox; + char mailbox[MAILBOX_NAME_SIZE]; // string representation of the id } s_finger_t, *finger_t; -/** +/* * Node data. */ -typedef struct node { +typedef struct s_node { int id; // my id - char* mailbox; // my usual mailbox name - s_finger_t fingers[NB_BITS]; // finger table (fingers[0] is my successor) + char mailbox[MAILBOX_NAME_SIZE]; // my mailbox name (string representation of the id) + s_finger_t *fingers; // finger table, of size nb_bits (fingers[0] is my successor) int pred_id; // predecessor id - char* pred_mailbox; // predecessor's mailbox name + char pred_mailbox[MAILBOX_NAME_SIZE]; // predecessor's mailbox name int next_finger_to_fix; // index of the next finger to fix in fix_fingers() - xbt_dynar_t comms; // current communications pending + msg_comm_t comm_receive; // current communication to receive + double last_change_date; // last time I changed a finger or my predecessor + RngStream stream; //RngStream for } s_node_t, *node_t; /** - * Task data + * Types of tasks exchanged between nodes. */ -typedef struct task_data { - int request_id; - int request_finger; - int answer_id; - const char* answer_to; - const char* issuer_host_name; // used for logging - int successor_id; // used when quitting - int pred_id; // used when quitting - // FIXME: remove successor_id and pred_id, request_id is enough +typedef enum { + TASK_FIND_SUCCESSOR, + TASK_FIND_SUCCESSOR_ANSWER, + TASK_GET_PREDECESSOR, + TASK_GET_PREDECESSOR_ANSWER, + TASK_NOTIFY, + TASK_SUCCESSOR_LEAVING, + TASK_PREDECESSOR_LEAVING, + TASK_PREDECESSOR_ALIVE, + TASK_PREDECESSOR_ALIVE_ANSWER +} e_task_type_t; + +/* + * Data attached with the tasks sent and received + */ +typedef struct s_task_data { + e_task_type_t type; // type of task + int request_id; // id paramater (used by some types of tasks) + int request_finger; // finger parameter (used by some types of tasks) + int answer_id; // answer (used by some types of tasks) + char answer_to[MAILBOX_NAME_SIZE]; // mailbox to send an answer to (if any) + const char* issuer_host_name; // used for logging } s_task_data_t, *task_data_t; -static int powers2[NB_BITS]; +static int *powers2; // utility functions static void chord_initialize(void); +static void chord_exit(void); static int normalize(int id); static int is_in_interval(int id, int start, int end); -static char* get_mailbox(int host_id); +static void get_mailbox(int host_id, char* mailbox); +static void task_free(void* task); static void print_finger_table(node_t node); static void set_finger(node_t node, int finger_index, int id); static void set_predecessor(node_t node, int predecessor_id); // process functions static int node(int argc, char *argv[]); +static void handle_task(node_t node, msg_task_t task); // Chord core static void create(node_t node); -static void join(node_t node, int known_id); +static int join(node_t node, int known_id); static void leave(node_t node); static int find_successor(node_t node, int id); static int remote_find_successor(node_t node, int ask_to_id, int id); @@ -77,57 +120,46 @@ static void notify(node_t node, int predecessor_candidate_id); static void remote_notify(node_t node, int notify_to, int predecessor_candidate_id); static void fix_fingers(node_t node); static void check_predecessor(node_t node); -static void quit_notify(node_t node, int to); - -// not implemented yet -//static void remote_move_keys(node_t node, int take_from_id); - -// deprecated -/* -static void bootstrap(node_t node, int known_id); -static int find_predecessor(node_t node, int id); -static int remote_find_predecessor(node_t node, int ask_to_id, int id); -static int remote_closest_preceding_finger(int ask_to_id, int id); -static void notify_predecessors(node_t node); -static void initialize_finger_table(node_t data, int known_id); -static void update_finger_table(node_t node, int candidate_id, int finger_index); -static void remote_update_finger_table(node_t node, int ask_to_id, int candidate_id, int finger_index); -static void notify_node_joined(node_t node, int new_node_id); -static void remote_notify_node_joined(node_t node, int notify_id); -*/ +static void random_lookup(node_t); +static void quit_notify(node_t node); +/** + * \brief Global initialization of the Chord simulation. + */ static void chord_initialize(void) { // compute the powers of 2 once for all + powers2 = xbt_new(int, nb_bits); int pow = 1; int i; - for (i = 0; i < NB_BITS; i++) { + for (i = 0; i < nb_bits; i++) { powers2[i] = pow; pow = pow << 1; } + nb_keys = pow; + XBT_DEBUG("Sets nb_keys to %d", nb_keys); +} + +static void chord_exit(void) +{ + xbt_free(powers2); } /** - * \brief Turns an id into an equivalent id in [0, NB_KEYS[ + * \brief Turns an id into an equivalent id in [0, nb_keys). * \param id an id * \return the corresponding normalized id */ static int normalize(int id) { - // make sure id >= 0 - while (id < 0) { - id += NB_KEYS; - } - // make sure id < NB_KEYS - id = id % NB_KEYS; - - return id; + // like id % nb_keys, but works with negatives numbers (and faster) + return id & (nb_keys - 1); } /** - * \brief Returns whether a id belongs to the interval [start, end]. + * \brief Returns whether an id belongs to the interval [start, end]. * - * The parameters are noramlized to make sure they are between 0 and CHORD_NB_KEYS - 1). + * The parameters are noramlized to make sure they are between 0 and nb_keys - 1). * 1 belongs to [62, 3] * 1 does not belong to [3, 62] * 63 belongs to [62, 3] @@ -148,11 +180,11 @@ static int is_in_interval(int id, int start, int end) // make sure end >= start and id >= start if (end < start) { - end += NB_KEYS; + end += nb_keys; } if (id < start) { - id += NB_KEYS; + id += nb_keys; } return id <= end; @@ -161,12 +193,25 @@ static int is_in_interval(int id, int start, int end) /** * \brief Gets the mailbox name of a host given its chord id. * \param node_id id of a node - * \return the name of its mailbox - * FIXME: free the memory + * \param mailbox pointer to where the mailbox name should be written + * (there must be enough space) */ -static char* get_mailbox(int node_id) +static void get_mailbox(int node_id, char* mailbox) { - return bprintf("mailbox%d", node_id); + snprintf(mailbox, MAILBOX_NAME_SIZE - 1, "%d", node_id); +} + +/** + * \brief Frees the memory used by a task. + * \param task the MSG task to destroy + */ +static void task_free(void* task) +{ + // TODO add a parameter data_free_function to MSG_task_create? + if(task != NULL){ + xbt_free(MSG_task_get_data(task)); + MSG_task_destroy(task); + } } /** @@ -175,29 +220,31 @@ static char* get_mailbox(int node_id) */ static void print_finger_table(node_t node) { - int i; - int pow = 1; - INFO0("My finger table:"); - INFO0("Start | Succ "); - for (i = 0; i < NB_BITS; i++) { - INFO2(" %3d | %3d ", (node->id + pow) % NB_KEYS, node->fingers[i].id); - pow = pow << 1; + if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) { + int i; + XBT_VERB("My finger table:"); + XBT_VERB("Start | Succ "); + for (i = 0; i < nb_bits; i++) { + XBT_VERB(" %3d | %3d ", (node->id + powers2[i]) % nb_keys, node->fingers[i].id); + } + XBT_VERB("Predecessor: %d", node->pred_id); } - INFO1("Predecessor: %d", node->pred_id); } /** * \brief Sets a finger of the current node. * \param node the current node - * \param finger_index index of the finger to set (0 to NB_BITS - 1) + * \param finger_index index of the finger to set (0 to nb_bits - 1) * \param id the id to set for this finger */ static void set_finger(node_t node, int finger_index, int id) { - node->fingers[finger_index].id = id; - xbt_free(node->fingers[finger_index].mailbox); - node->fingers[finger_index].mailbox = get_mailbox(id); - INFO2("My new finger #%d is %d", finger_index, id); + if (id != node->fingers[finger_index].id) { + node->fingers[finger_index].id = id; + get_mailbox(id, node->fingers[finger_index].mailbox); + node->last_change_date = MSG_get_clock(); + XBT_DEBUG("My new finger #%d is %d", finger_index, id); + } } /** @@ -207,14 +254,16 @@ static void set_finger(node_t node, int finger_index, int id) */ static void set_predecessor(node_t node, int predecessor_id) { - node->pred_id = predecessor_id; - xbt_free(node->pred_mailbox); + if (predecessor_id != node->pred_id) { + node->pred_id = predecessor_id; - if (predecessor_id != -1) { - node->pred_mailbox = get_mailbox(predecessor_id); - } + if (predecessor_id != -1) { + get_mailbox(predecessor_id, node->pred_mailbox); + } + node->last_change_date = MSG_get_clock(); - INFO1("My new predecessor is %d", predecessor_id); + XBT_DEBUG("My new predecessor is %d", predecessor_id); + } } /** @@ -226,194 +275,267 @@ static void set_predecessor(node_t node, int predecessor_id) */ int node(int argc, char *argv[]) { + + /* Reduce the run size for the MC */ + if(MC_is_active()){ + periodic_stabilize_delay = 8; + periodic_fix_fingers_delay = 8; + periodic_check_predecessor_delay = 8; + } + double init_time = MSG_get_clock(); - m_task_t task = NULL; - msg_comm_t comm_send = NULL; - msg_comm_t comm_receive = NULL; + msg_task_t task_received = NULL; int i; - char* mailbox = NULL; + int join_success = 0; double deadline; - double next_stabilize_date = init_time + 10; - double next_fix_fingers_date = init_time + 10; - double next_check_predecessor_date = init_time + 10; + double next_stabilize_date = init_time + periodic_stabilize_delay; + double next_fix_fingers_date = init_time + periodic_fix_fingers_delay; + double next_check_predecessor_date = init_time + periodic_check_predecessor_delay; + double next_lookup_date = init_time + periodic_lookup_delay; + + #ifdef HAVE_MC + int listen = 0; + int no_op = 0; + int sub_protocol = 0; + #endif - xbt_assert0(argc == 3 || argc == 5, "Wrong number of arguments for this node"); + xbt_assert(argc == 3 || argc == 5, "Wrong number of arguments for this node"); // initialize my node s_node_t node = {0}; node.id = atoi(argv[1]); - node.mailbox = get_mailbox(node.id); + get_mailbox(node.id, node.mailbox); node.next_finger_to_fix = 0; - node.comms = xbt_dynar_new(sizeof(msg_comm_t), NULL); + node.fingers = xbt_new0(s_finger_t, nb_bits); + node.last_change_date = init_time; - for (i = 0; i < NB_BITS; i++) { + for (i = 0; i < nb_bits; i++) { + node.fingers[i].id = -1; set_finger(&node, i, node.id); } if (argc == 3) { // first ring deadline = atof(argv[2]); create(&node); + join_success = 1; } else { int known_id = atoi(argv[2]); - double sleep_time = atof(argv[3]); + //double sleep_time = atof(argv[3]); deadline = atof(argv[4]); + /* // sleep before starting - INFO1("Let's sleep during %f", sleep_time); + XBT_DEBUG("Let's sleep during %f", sleep_time); MSG_process_sleep(sleep_time); - INFO0("Hey! Let's join the system."); + */ + XBT_DEBUG("Hey! Let's join the system."); - join(&node, known_id); + join_success = join(&node, known_id); } - while (MSG_get_clock() < init_time + deadline) { + if (join_success) { + while (MSG_get_clock() < init_time + deadline +// && MSG_get_clock() < node.last_change_date + 1000 + && MSG_get_clock() < max_simulation_time) { + + if (node.comm_receive == NULL) { + task_received = NULL; + node.comm_receive = MSG_task_irecv(&task_received, node.mailbox); + // FIXME: do not make MSG_task_irecv() calls from several functions + } - // see if some communications are finished - int index; - while ((index = MSG_comm_testany(node.comms)) != -1) { - comm_send = xbt_dynar_get_as(node.comms, index, msg_comm_t); - xbt_dynar_remove_at(node.comms, index, &comm_send); - MSG_comm_destroy(comm_send); + //XBT_INFO("Node %d is ring member : %d", node.id, is_ring_member(known_id, node.id) != -1); + + if (!MSG_comm_test(node.comm_receive)) { + + // no task was received: make some periodic calls + +#ifdef HAVE_MC + if(MC_is_active()){ + if(!MC_visited_reduction() && no_op){ + MC_cut(); + } + if(listen == 0 && (sub_protocol = MC_random(0, 4)) > 0){ + if(sub_protocol == 1) + stabilize(&node); + else if(sub_protocol == 2) + fix_fingers(&node); + else if(sub_protocol == 3) + check_predecessor(&node); + else + random_lookup(&node); + listen = 1; + }else{ + MSG_process_sleep(5); + if(!MC_visited_reduction()) + no_op = 1; + } + }else{ + if (MSG_get_clock() >= next_stabilize_date) { + stabilize(&node); + next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay; + }else if (MSG_get_clock() >= next_fix_fingers_date) { + fix_fingers(&node); + next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay; + }else if (MSG_get_clock() >= next_check_predecessor_date) { + check_predecessor(&node); + next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay; + }else if (MSG_get_clock() >= next_lookup_date) { + random_lookup(&node); + next_lookup_date = MSG_get_clock() + periodic_lookup_delay; + }else { + // nothing to do: sleep for a while + MSG_process_sleep(5); + } + } +#else + if (MSG_get_clock() >= next_stabilize_date) { + stabilize(&node); + next_stabilize_date = MSG_get_clock() + periodic_stabilize_delay; + }else if (MSG_get_clock() >= next_fix_fingers_date) { + fix_fingers(&node); + next_fix_fingers_date = MSG_get_clock() + periodic_fix_fingers_delay; + }else if (MSG_get_clock() >= next_check_predecessor_date) { + check_predecessor(&node); + next_check_predecessor_date = MSG_get_clock() + periodic_check_predecessor_delay; + }else if (MSG_get_clock() >= next_lookup_date) { + random_lookup(&node); + next_lookup_date = MSG_get_clock() + periodic_lookup_delay; + }else { + // nothing to do: sleep for a while + MSG_process_sleep(5); + } +#endif + + } else { + // a transfer has occurred + + msg_error_t status = MSG_comm_get_status(node.comm_receive); + + if (status != MSG_OK) { + XBT_DEBUG("Failed to receive a task. Nevermind."); + MSG_comm_destroy(node.comm_receive); + node.comm_receive = NULL; + } + else { + // the task was successfully received + MSG_comm_destroy(node.comm_receive); + node.comm_receive = NULL; + handle_task(&node, task_received); + } + } } - if (comm_receive == NULL) { - task = NULL; - comm_receive = MSG_task_irecv(&task, node.mailbox); + if (node.comm_receive) { + MSG_comm_destroy(node.comm_receive); + node.comm_receive = NULL; } - if (!MSG_comm_test(comm_receive)) { + // leave the ring + leave(&node); + } - // no task was received: make some periodic calls - if (MSG_get_clock() >= next_stabilize_date) { - stabilize(&node); - next_stabilize_date = MSG_get_clock() + 10; - } - else if (MSG_get_clock() >= next_fix_fingers_date) { - fix_fingers(&node); - next_fix_fingers_date = MSG_get_clock() + 10; - } - else if (MSG_get_clock() >= next_check_predecessor_date) { - check_predecessor(&node); - next_check_predecessor_date = MSG_get_clock() + 10; - } - else { - // nothing to do: sleep for a while - MSG_process_sleep(5); - } - } - else { - // a task was received + // stop the simulation + xbt_free(node.fingers); + return 0; +} - MSG_comm_destroy(comm_receive); - comm_receive = NULL; - const char* task_name = MSG_task_get_name(task); - task_data_t task_data = (task_data_t) MSG_task_get_data(task); +/** + * \brief This function is called when the current node receives a task. + * \param node the current node + * \param task the task to handle (don't touch it then: + * it will be destroyed, reused or forwarded) + */ +static void handle_task(node_t node, msg_task_t task) { - if (!strcmp(task_name, "Find Successor")) { - INFO2("Receiving a 'Find Successor' Request from %s for id %d", - task_data->issuer_host_name, task_data->request_id); - // is my successor the successor? - if (is_in_interval(task_data->request_id, node.id + 1, node.fingers[0].id)) { - task_data->answer_id = node.fingers[0].id; - MSG_task_set_name(task, "Find Successor Answer"); - INFO3("Sending back a 'Find Successor Answer' to %s: the successor of %d is %d", - task_data->issuer_host_name, task_data->request_id, task_data->answer_id); - comm_send = MSG_task_isend(task, task_data->answer_to); - xbt_dynar_push(node.comms, &comm_send); - } - else { - // otherwise, forward the request to the closest preceding finger in my table - int closest = closest_preceding_node(&node, task_data->request_id); - INFO2("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d", - task_data->request_id, closest); - mailbox = get_mailbox(closest); - comm_send = MSG_task_isend(task, mailbox); - xbt_dynar_push(node.comms, &comm_send); - xbt_free(mailbox); - } - } + XBT_DEBUG("Handling task %p", task); + char mailbox[MAILBOX_NAME_SIZE]; + task_data_t task_data = (task_data_t) MSG_task_get_data(task); + e_task_type_t type = task_data->type; - else if (!strcmp(task_name, "Get Predecessor")) { - INFO1("Receiving a 'Get Predecessor' Request from %s", task_data->issuer_host_name); - task_data->answer_id = node.pred_id; - MSG_task_set_name(task, "Get Predecessor Answer"); - INFO2("Sending back a 'Get Predecessor Answer' to %s: my predecessor is %d", - task_data->issuer_host_name, task_data->answer_id); - comm_send = MSG_task_isend(task, task_data->answer_to); - xbt_dynar_push(node.comms, &comm_send); - } - /* - else if (!strcmp(task_name, "Find Predecessor")) { - INFO2("Receiving a 'Find Predecessor' Request from %s for id %d", task_data->issuer_host_name, task_data->request_id); - // am I the predecessor? - if (is_in_interval(task_data->request_id, node.id + 1, node.fingers[0].id)) { - task_data->answer_id = node.id; - MSG_task_set_name(task, "Find Predecessor Answer"); - INFO3("Sending back a 'Find Predecessor' Answer to %s: the predecessor of %d is %d", task_data->issuer_host_name, task_data->request_id, task_data->answer_id); - comm = MSG_task_isend(task, task_data->answer_to); - xbt_dynar_push(node.comms, &comm); - } - else { - // otherwise, forward the request to the closest preceding finger in my table - int closest = closest_preceding_node(&node, task_data->request_id); - INFO2("Forwarding 'Find Predecessor' request for id %d to my closest preceding finger %d", task_data->request_id, closest); - mailbox = get_mailbox(closest); - comm = MSG_task_isend(task, mailbox); - xbt_dynar_push(node.comms, &comm); - xbt_free(mailbox); - } - } - */ + switch (type) { - /* - else if (!strcmp(task_name, "Notify Node Joined")) { - // someone may be my new neighboor - INFO1("Receiving a 'Notify Node Joine' request from %s", task_data->issuer_host_name); - notify_node_joined(&node, task_data->request_id); - } - else if (!strcmp(task_name, "Update Finger")) { - // someone is telling me that he may be my new finger - INFO1("Receiving an 'Update Finger' request from %s", task_data->issuer_host_name); - update_finger_table(&node, task_data->request_id, task_data->request_finger); - } - */ - else if (!strcmp(task_name, "Notify")) { - // someone is telling me that he may be my new predecessor - INFO1("Receiving a 'Notify' request from %s", task_data->issuer_host_name); - notify(&node, task_data->request_id); - } - else if (!strcmp(task_name, "Predecessor Leaving")) { - // my predecessor is about to quit - INFO1("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name); - // modify my predecessor - set_predecessor(&node, task_data->pred_id); - /*TODO : - >> notify my new predecessor - >> send a notify_predecessors !! - */ - } - else if (!strcmp(task_name, "Successor Leaving")) { - // my successor is about to quit - INFO1("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name); - // modify my successor FIXME : this should be implicit ? - set_finger(&node, 0, task_data->successor_id); - /* TODO - >> notify my new successor - >> update my table & predecessors table */ - } + case TASK_FIND_SUCCESSOR: + XBT_DEBUG("Receiving a 'Find Successor' request from %s for id %d", + task_data->issuer_host_name, task_data->request_id); + // is my successor the successor? + if (is_in_interval(task_data->request_id, node->id + 1, node->fingers[0].id)) { + task_data->type = TASK_FIND_SUCCESSOR_ANSWER; + task_data->answer_id = node->fingers[0].id; + XBT_DEBUG("Sending back a 'Find Successor Answer' to %s (mailbox %s): the successor of %d is %d", + task_data->issuer_host_name, + task_data->answer_to, + task_data->request_id, task_data->answer_id); + MSG_task_dsend(task, task_data->answer_to, task_free); } - } + else { + // otherwise, forward the request to the closest preceding finger in my table + int closest = closest_preceding_node(node, task_data->request_id); + XBT_DEBUG("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d", + task_data->request_id, closest); + get_mailbox(closest, mailbox); + MSG_task_dsend(task, mailbox, task_free); + } + break; + + case TASK_GET_PREDECESSOR: + XBT_DEBUG("Receiving a 'Get Predecessor' request from %s", task_data->issuer_host_name); + task_data->type = TASK_GET_PREDECESSOR_ANSWER; + task_data->answer_id = node->pred_id; + XBT_DEBUG("Sending back a 'Get Predecessor Answer' to %s via mailbox '%s': my predecessor is %d", + task_data->issuer_host_name, + task_data->answer_to, task_data->answer_id); + MSG_task_dsend(task, task_data->answer_to, task_free); + break; + + case TASK_NOTIFY: + // someone is telling me that he may be my new predecessor + XBT_DEBUG("Receiving a 'Notify' request from %s", task_data->issuer_host_name); + notify(node, task_data->request_id); + task_free(task); + break; + + case TASK_PREDECESSOR_LEAVING: + // my predecessor is about to quit + XBT_DEBUG("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name); + // modify my predecessor + set_predecessor(node, task_data->request_id); + task_free(task); + /*TODO : + >> notify my new predecessor + >> send a notify_predecessors !! + */ + break; + + case TASK_SUCCESSOR_LEAVING: + // my successor is about to quit + XBT_DEBUG("Receiving a 'Successor Leaving' message from %s", task_data->issuer_host_name); + // modify my successor FIXME : this should be implicit ? + set_finger(node, 0, task_data->request_id); + task_free(task); + /* TODO + >> notify my new successor + >> update my table & predecessors table */ + break; + + case TASK_FIND_SUCCESSOR_ANSWER: + case TASK_GET_PREDECESSOR_ANSWER: + case TASK_PREDECESSOR_ALIVE_ANSWER: + XBT_DEBUG("Ignoring unexpected task of type %d (%p)", (int)type, task); + task_free(task); + break; + + case TASK_PREDECESSOR_ALIVE: + XBT_DEBUG("Receiving a 'Predecessor Alive' request from %s", task_data->issuer_host_name); + task_data->type = TASK_PREDECESSOR_ALIVE_ANSWER; + XBT_DEBUG("Sending back a 'Predecessor Alive Answer' to %s (mailbox %s)", + task_data->issuer_host_name, + task_data->answer_to); + MSG_task_dsend(task, task_data->answer_to, task_free); + break; - // leave the ring and stop the simulation - leave(&node); - xbt_dynar_free(&node.comms); - xbt_free(node.mailbox); - xbt_free(node.pred_mailbox); - for (i = 0; i < NB_BITS - 1; i++) { - xbt_free(node.fingers[i].mailbox); } - return 0; } /** @@ -422,7 +544,7 @@ int node(int argc, char *argv[]) */ static void create(node_t node) { - INFO0("Create a new Chord ring..."); + XBT_DEBUG("Create a new Chord ring..."); set_predecessor(node, -1); // -1 means that I have no predecessor print_finger_table(node); } @@ -432,93 +554,31 @@ static void create(node_t node) * already in the ring * \param node the current node * \param known_id id of a node already in the ring + * \return 1 if the join operation succeeded, 0 otherwise */ -static void join(node_t node, int known_id) +static int join(node_t node, int known_id) { - INFO2("Joining the ring with id %d, knowing node %d", node->id, known_id); + XBT_INFO("Joining the ring with id %d, knowing node %d", node->id, known_id); set_predecessor(node, -1); // no predecessor (yet) - int successor_id; - do { - successor_id = remote_find_successor(node, known_id, node->id); - INFO0("My 'Find Successor' request has failed, let's try again"); - } while (successor_id == -1); - - set_finger(node, 0, successor_id); - print_finger_table(node); -} - -/** - * \brief Notifies the current node that a node has just joined the network. - * \param node the current node - * \param new_node_id id of the new node in the network - */ -/* -static void notify_node_joined(node_t node, int new_node_id) -{ - INFO1("A new node %d has joined the network, let's see if it is a neighboor", new_node_id); - if (is_in_interval(new_node_id, node->id + 1, node->fingers[0].id - 1)) { - // new_node_id is my new successor - set_finger(node, 0, new_node_id); - bootstrap(node, new_node_id); + /* + int i; + for (i = 0; i < nb_bits; i++) { + set_finger(node, i, known_id); } + */ - if (is_in_interval(new_node_id, node->pred_id + 1, node->id - 1)) { - // new_node_id is my new predecessor - set_predecessor(node, new_node_id); - bootstrap(node, new_node_id); + int successor_id = remote_find_successor(node, known_id, node->id); + if (successor_id == -1) { + XBT_INFO("Cannot join the ring."); } -} -*/ - -/** - * \brief Notifies a remote node that the current node has just joined - * the network. - * \param node the current node - * \param notify_id id of the remote node to notify - */ -/* -static void remote_notify_node_joined(node_t node, int notify_id) -{ - task_data_t req_data = xbt_new0(s_task_data_t, 1); - req_data->request_id = node->id; - req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); - - // send a "Notify" request to notify_id - INFO2("Sending a 'Notify' request to %d because I have joined the network with id %d", notify_id, node->id); - m_task_t task = MSG_task_create("Notify", 1000, 5000, req_data); - char* mailbox = get_mailbox(notify_id); - msg_comm_t comm = MSG_task_isend(task, mailbox); - xbt_dynar_push(node->comms, &comm); - xbt_free(mailbox); -} -*/ - -/** -* \brief Let the current node send queries to fill in its own finger table. -* \param node the current node -* \param ask_to_id id of a node to send queries to -*/ -/* -static void bootstrap(node_t node, int ask_to_id) -{ - INFO0("Filling my finger table"); - int i, pred_id, succ_id; - int pow = 1; - - for (i = 0; i < NB_BITS; i++) - { - pred_id = remote_find_successor(node, ask_to_id, pow); - do { - succ_id = pred_id; - pred_id = remote_find_predecessor(node, pred_id, pred_id); - } while (pred_id >= pow); - - pow = pow << 1; - set_finger(node, i, succ_id); + else { + set_finger(node, 0, successor_id); + print_finger_table(node); } + + return successor_id != -1; } -*/ /** * \brief Makes the current node quit the system @@ -526,168 +586,54 @@ static void bootstrap(node_t node, int ask_to_id) */ static void leave(node_t node) { - INFO0("Well Guys! I Think it's time for me to quit ;)"); - quit_notify(node, 1); // notify to my successor ( >>> 1 ); - quit_notify(node, -1); // notify my predecessor ( >>> -1); - // TODO ... + XBT_DEBUG("Well Guys! I Think it's time for me to quit ;)"); + quit_notify(node); + RngStream_DeleteStream(&node->stream); } -/* - * \brief Notifies the successor or the predecessor of the current node +/** + * \brief Notifies the successor and the predecessor of the current node * of the departure * \param node the current node - * \param to 1 to notify the successor, -1 to notify the predecessor - * FIXME: notify both nodes with only one call */ -static void quit_notify(node_t node, int to) +static void quit_notify(node_t node) { - task_data_t req_data = xbt_new0(s_task_data_t, 1); - req_data->request_id = node->id; - req_data->successor_id = node->fingers[0].id; - req_data->pred_id = node->pred_id; + char mailbox[MAILBOX_NAME_SIZE]; + //send the PREDECESSOR_LEAVING to our successor + task_data_t req_data = xbt_new0(s_task_data_t,1); + req_data->type = TASK_PREDECESSOR_LEAVING; + req_data->request_id = node->pred_id; + get_mailbox(node->id, req_data->answer_to); req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); - const char *task_name = NULL; - const char* to_mailbox = NULL; - if (to == 1) // notify my successor - { - to_mailbox = node->fingers[0].mailbox; - INFO2("Telling my Successor %d about my departure via mailbox %s", - node->fingers[0].id, to_mailbox); - task_name = "Predecessor Leaving"; - - } - else if (to == -1) // notify my predecessor - { - if (node->pred_id == -1) { - return; - } - - to_mailbox = node->pred_mailbox; - INFO2("Telling my Predecessor %d about my departure via mailbox %s", - node->pred_id, to_mailbox); - task_name = "Predecessor Leaving"; + msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data); + XBT_DEBUG("Sending a 'PREDECESSOR_LEAVING' to my successor %d",node->fingers[0].id); + if (MSG_task_send_with_timeout(task_sent, node->fingers[0].mailbox, timeout)== + MSG_TIMEOUT) { + XBT_DEBUG("Timeout expired when sending a 'PREDECESSOR_LEAVING' to my successor %d", + node->fingers[0].id); + task_free(task_sent); } - m_task_t task = MSG_task_create(task_name, 1000, 5000, req_data); - //char* mailbox = get_mailbox(to_mailbox); - msg_comm_t comm = MSG_task_isend(task, to_mailbox); - xbt_dynar_push(node->comms, &comm); -} - -/* - * \brief Initializes my finger table, knowing the id of a node already in the system. - * \param node the current node - * \param known_id id of a node already in the system - */ -/* -static void initialize_finger_table(node_t node, int known_id) -{ - int my_id = node->id; - int i; - int pow = 1; // 2^i - - INFO0("Initializing my finger table..."); - - // ask known_id who is my immediate successor - node->fingers[0].id = remote_find_successor(node, known_id, my_id + 1); - node->fingers[0].mailbox = get_mailbox(node->fingers[0].id); - - // find all other fingers - for (i = 0; i < NB_BITS - 1; i++) { - pow = pow << 1; // equivalent to pow = pow * 2 - if (is_in_interval(my_id + pow, my_id, node->fingers[i].id - 1)) { - // I already have the info for this finger - node->fingers[i + 1].id = node->fingers[i].id; - } - else { - // I don't have the info, ask the only guy I know - node->fingers[i + 1].id = remote_find_successor(node, known_id, my_id + pow); - } - node->fingers[i + 1].mailbox = get_mailbox(node->fingers[i + 1].id); - } - - node->pred_id = find_predecessor(node, node->id); - node->pred_mailbox = get_mailbox(node->pred_id); - - INFO0("Finger table initialized!"); - print_finger_table(node); -} -*/ - -/** - * \brief Notifies some nodes that the current node may have became their finger. - * \param node the current node, which has just joined the system - */ -/* -static void notify_predecessors(node_t node) -{ - int i, pred_id; - int pow = 1; - for (i = 0; i < NB_BITS; i++) { - // find the closest node whose finger #i can be me - pred_id = find_predecessor(node, node->id - pow + 1); // note: no "+1" in the article! - if (pred_id != node->id) { - remote_update_finger_table(node, pred_id, node->id, i); - } - pow = pow << 1; // pow = pow * 2 + //send the SUCCESSOR_LEAVING to our predecessor + get_mailbox(node->pred_id, mailbox); + task_data_t req_data_s = xbt_new0(s_task_data_t,1); + req_data_s->type = TASK_SUCCESSOR_LEAVING; + req_data_s->request_id = node->fingers[0].id; + req_data_s->request_id = node->pred_id; + get_mailbox(node->id, req_data_s->answer_to); + req_data_s->issuer_host_name = MSG_host_get_name(MSG_host_self()); + + msg_task_t task_sent_s = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data_s); + XBT_DEBUG("Sending a 'SUCCESSOR_LEAVING' to my predecessor %d",node->pred_id); + if (MSG_task_send_with_timeout(task_sent_s, mailbox, timeout)== + MSG_TIMEOUT) { + XBT_DEBUG("Timeout expired when sending a 'SUCCESSOR_LEAVING' to my predecessor %d", + node->pred_id); + task_free(task_sent_s); } -} -*/ -/** - * \brief Tells the current node that a node may have became its new finger. - * \param node the current node - * \param candidate_id id of the node that may be a new finger of the current node - * \param finger_index index of the finger to update - */ -/* -static void update_finger_table(node_t node, int candidate_id, int finger_index) -{ - int pow = 1; - int i; - for (i = 0; i < finger_index; i++) { - pow = pow << 1; - } - - // if (is_in_interval(candidate_id, node->id + pow, node->fingers[finger_index].id - 1)) { - if (is_in_interval(candidate_id, node->id, node->fingers[finger_index].id - 1)) { -// INFO3("Candidate %d is between %d and %d!", candidate_id, node->id + pow, node->fingers[finger_index].id - 1); - // candidate_id is my new finger - set_finger(node, finger_index, candidate_id); - print_finger_table(node); - - if (node->pred_id != node->id) { // FIXME: is this necessary? - // my predecessor may be concerned too - remote_update_finger_table(node, node->pred_id, candidate_id, finger_index); - } - } } -*/ - -/** - * \brief Tells a remote node that a node may have became its new finger. - * \param ask_to_id id of the remote node to update - * \param candidate_id id of the node that may be a new finger of the remote node - * \param finger_index index of the finger to update - */ -/* -static void remote_update_finger_table(node_t node, int ask_to_id, int candidate_id, int finger_index) -{ - task_data_t req_data = xbt_new0(s_task_data_t, 1); - req_data->request_id = candidate_id; - req_data->request_finger = finger_index; - req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); - - // send a "Update Finger" request to ask_to_id - INFO3("Sending an 'Update Finger' request to %d: his finger #%d may be %d now", ask_to_id, finger_index, candidate_id); - m_task_t task = MSG_task_create("Update Finger", 1000, 5000, req_data); - char* mailbox = get_mailbox(ask_to_id); - msg_comm_t comm = MSG_task_isend(task, mailbox); - xbt_dynar_push(node->comms, &comm); - xbt_free(mailbox); -} -*/ /** * \brief Makes the current node find the successor node of an id. @@ -717,32 +663,85 @@ static int find_successor(node_t node, int id) static int remote_find_successor(node_t node, int ask_to, int id) { int successor = -1; - s_task_data_t req_data; - char* mailbox = bprintf("%s Find Successor", node->mailbox); - req_data.request_id = id; - req_data.answer_to = mailbox; - req_data.issuer_host_name = MSG_host_get_name(MSG_host_self()); + int stop = 0; + char mailbox[MAILBOX_NAME_SIZE]; + get_mailbox(ask_to, mailbox); + task_data_t req_data = xbt_new0(s_task_data_t, 1); + req_data->type = TASK_FIND_SUCCESSOR; + req_data->request_id = id; + get_mailbox(node->id, req_data->answer_to); + req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); // send a "Find Successor" request to ask_to_id - INFO2("Sending a 'Find Successor' request to %d for key %d", ask_to, id); - m_task_t task = MSG_task_create("Find Successor", 1000, 5000, &req_data); - MSG_error_t res = MSG_task_send_with_timeout(task, get_mailbox(ask_to), 20); - if (res == MSG_OK) { + msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data); + XBT_DEBUG("Sending a 'Find Successor' request (task %p) to %d for id %d", task_sent, ask_to, id); + msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout); + + if (res != MSG_OK) { + XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d", + task_sent, ask_to, id); + task_free(task_sent); + } + else { // receive the answer - task = NULL; - res = MSG_task_receive_with_timeout(&task, req_data.answer_to, 20); + XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer", + task_sent, ask_to, id); - if (res == MSG_OK) { + do { + if (node->comm_receive == NULL) { + msg_task_t task_received = NULL; + node->comm_receive = MSG_task_irecv(&task_received, node->mailbox); + } - task_data_t ans_data; - ans_data = MSG_task_get_data(task); - successor = ans_data->answer_id; - INFO2("Received the answer to my 'Find Successor' request: the successor of key %d is %d", id, successor); - } + res = MSG_comm_wait(node->comm_receive, timeout); + + if (res != MSG_OK) { + XBT_DEBUG("Failed to receive the answer to my 'Find Successor' request (task %p): %d", + task_sent, (int)res); + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + } + else { + msg_task_t task_received = MSG_comm_get_task(node->comm_receive); + XBT_DEBUG("Received a task (%p)", task_received); + task_data_t ans_data = MSG_task_get_data(task_received); + + // Once upon a time, our code assumed that here, task_received != task_sent all the time + // + // This assumption is wrong (as messages from differing round can interleave), leading to a bug in our code. + // We failed to find this bug directly, as it only occured on large platforms, leading to hardly usable traces. + // Instead, we used the model-checker to track down the issue by adding the following test here in the code: + // if (MC_is_active()) { + // MC_assert(task_received == task_sent); + // } + // That explained the bug in a snap, with a very cool example and everything. + // + // This MC_assert is now desactivated as the case is now properly handled in our code and we don't want the + // MC to fail any further under that condition, but this comment is here to as a memorial for this first + // brillant victory of the model-checking in the SimGrid community :) + + if (task_received != task_sent) { + // this is not the expected answer + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + handle_task(node, task_received); + } + else { + // this is our answer + XBT_DEBUG("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d", + ans_data->request_id, task_received, id, ans_data->answer_id); + successor = ans_data->answer_id; + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + task_free(task_received); + } + } + } while (!stop); } - xbt_free(mailbox); return successor; } @@ -756,90 +755,73 @@ static int remote_find_successor(node_t node, int ask_to, int id) static int remote_get_predecessor(node_t node, int ask_to) { int predecessor_id = -1; - s_task_data_t req_data; - char* mailbox = bprintf("%s Get Predecessor", node->mailbox); - req_data.answer_to = mailbox; - req_data.issuer_host_name = MSG_host_get_name(MSG_host_self()); + int stop = 0; + char mailbox[MAILBOX_NAME_SIZE]; + get_mailbox(ask_to, mailbox); + task_data_t req_data = xbt_new0(s_task_data_t, 1); + req_data->type = TASK_GET_PREDECESSOR; + get_mailbox(node->id, req_data->answer_to); + req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); // send a "Get Predecessor" request to ask_to_id - INFO1("Sending a 'Get Predecessor' request to %d", ask_to); - m_task_t task = MSG_task_create("Get Predecessor", 1000, 5000, &req_data); - MSG_error_t res = MSG_task_send_with_timeout(task, get_mailbox(ask_to), 20); - - if (res == MSG_OK) { + XBT_DEBUG("Sending a 'Get Predecessor' request to %d", ask_to); + msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data); + msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout); + + if (res != MSG_OK) { + XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d", + task_sent, ask_to); + task_free(task_sent); + } + else { // receive the answer - task = NULL; - res = MSG_task_receive_with_timeout(&task, req_data.answer_to, 20); - - if (res == MSG_OK) { - - task_data_t ans_data; - ans_data = MSG_task_get_data(task); - predecessor_id = ans_data->answer_id; - INFO2("Received the answer to my 'Get Predecessor' request: the predecessor of node %d is %d", ask_to, predecessor_id); - } - } + XBT_DEBUG("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'", + task_sent, ask_to, req_data->answer_to); - xbt_free(mailbox); - return predecessor_id; -} + do { + if (node->comm_receive == NULL) { // FIXME simplify this + msg_task_t task_received = NULL; + node->comm_receive = MSG_task_irecv(&task_received, node->mailbox); + } -/** - * \brief Makes the current node find the predecessor node of an id. - * \param node the current node - * \param id the id to find - * \return the id of the predecessor node - */ -/* -static int find_predecessor(node_t node, int id) -{ - if (node->id == node->fingers[0].id) { - // I am the only node in the system - return node->id; - } + res = MSG_comm_wait(node->comm_receive, timeout); - if (is_in_interval(id, node->id + 1, node->fingers[0].id)) { - return node->id; + if (res != MSG_OK) { + XBT_DEBUG("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d", + task_sent, (int)res); + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + } + else { + msg_task_t task_received = MSG_comm_get_task(node->comm_receive); + task_data_t ans_data = MSG_task_get_data(task_received); + + /*if (MC_is_active()) { + MC_assert(task_received == task_sent); + }*/ + + if (task_received != task_sent) { + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + handle_task(node, task_received); + } + else { + XBT_DEBUG("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d", + task_received, ask_to, ans_data->answer_id); + predecessor_id = ans_data->answer_id; + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + task_free(task_received); + } + } + } while (!stop); } - int ask_to = closest_preceding_node(node, id); - return remote_find_predecessor(node, ask_to, id); -} -*/ -/** - * \brief Asks another node the predecessor node of an id. - * \param node the current node - * \param ask_to the node to ask to - * \param id the id to find - * \return the id of the predecessor node - */ -/* -static int remote_find_predecessor(node_t node, int ask_to, int id) -{ - s_task_data_t req_data; - char* mailbox = bprintf("%s Find Predecessor", node->mailbox); - req_data.request_id = id; - req_data.answer_to = mailbox; - req_data.issuer_host_name = MSG_host_get_name(MSG_host_self()); - - // send a "Find Predecessor" request to ask_to - INFO2("Sending a 'Find Predecessor' request to %d for key %d", ask_to, id); - m_task_t task = MSG_task_create("Find Predecessor", 1000, 5000, &req_data); - MSG_task_send(task, get_mailbox(ask_to)); - - // receive the answer - task = NULL; - MSG_task_receive(&task, req_data.answer_to); - task_data_t ans_data; - ans_data = MSG_task_get_data(task); - int predecessor = ans_data->answer_id; - xbt_free(mailbox); - INFO2("Received the answer to my 'Find Predecessor' request: the predecessor of key %d is %d", id, predecessor); - - return predecessor; + return predecessor_id; } -*/ /** * \brief Returns the closest preceding finger of an id @@ -851,7 +833,7 @@ static int remote_find_predecessor(node_t node, int ask_to, int id) int closest_preceding_node(node_t node, int id) { int i; - for (i = NB_BITS - 1; i >= 0; i--) { + for (i = nb_bits - 1; i >= 0; i--) { if (is_in_interval(node->fingers[i].id, node->id + 1, id - 1)) { return node->fingers[i].id; } @@ -866,7 +848,7 @@ int closest_preceding_node(node_t node, int id) */ static void stabilize(node_t node) { - INFO0("Stabilizing node"); + XBT_DEBUG("Stabilizing node"); // get the predecessor of my immediate successor int candidate_id; @@ -902,7 +884,7 @@ static void notify(node_t node, int predecessor_candidate_id) { print_finger_table(node); } else { - INFO1("I don't have to change my predecessor to %d", predecessor_candidate_id); + XBT_DEBUG("I don't have to change my predecessor to %d", predecessor_candidate_id); } } @@ -914,18 +896,18 @@ static void notify(node_t node, int predecessor_candidate_id) { */ static void remote_notify(node_t node, int notify_id, int predecessor_candidate_id) { - task_data_t req_data = xbt_new0(s_task_data_t, 1); - req_data->request_id = predecessor_candidate_id; - req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); - - // send a "Notify" request to notify_id - INFO1("Sending a 'Notify' request to %d", notify_id); - m_task_t task = MSG_task_create("Notify", 1000, 5000, req_data); - char* mailbox = get_mailbox(notify_id); - msg_comm_t comm = MSG_task_isend(task, mailbox); - xbt_dynar_push(node->comms, &comm); - xbt_free(mailbox); -} + task_data_t req_data = xbt_new0(s_task_data_t, 1); + req_data->type = TASK_NOTIFY; + req_data->request_id = predecessor_candidate_id; + req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); + + // send a "Notify" request to notify_id + msg_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data); + XBT_DEBUG("Sending a 'Notify' request (task %p) to %d", task, notify_id); + char mailbox[MAILBOX_NAME_SIZE]; + get_mailbox(notify_id, mailbox); + MSG_task_dsend(task, mailbox, task_free); + } /** * \brief This function is called periodically. @@ -934,7 +916,7 @@ static void remote_notify(node_t node, int notify_id, int predecessor_candidate_ */ static void fix_fingers(node_t node) { - INFO0("Fixing fingers"); + XBT_DEBUG("Fixing fingers"); int i = node->next_finger_to_fix; int id = find_successor(node, node->id + powers2[i]); if (id != -1) { @@ -943,7 +925,7 @@ static void fix_fingers(node_t node) { set_finger(node, i, id); print_finger_table(node); } - node->next_finger_to_fix = (i + 1) % NB_BITS; + node->next_finger_to_fix = (i + 1) % nb_bits; } } @@ -954,8 +936,87 @@ static void fix_fingers(node_t node) { */ static void check_predecessor(node_t node) { - INFO0("Checking whether my predecessor is alive"); - // TODO + XBT_DEBUG("Checking whether my predecessor is alive"); + + if(node->pred_id == -1) + return; + + int stop = 0; + + char mailbox[MAILBOX_NAME_SIZE]; + get_mailbox(node->pred_id, mailbox); + task_data_t req_data = xbt_new0(s_task_data_t,1); + req_data->type = TASK_PREDECESSOR_ALIVE; + req_data->request_id = node->pred_id; + get_mailbox(node->id, req_data->answer_to); + req_data->issuer_host_name = MSG_host_get_name(MSG_host_self()); + + msg_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data); + XBT_DEBUG("Sending a 'Predecessor Alive' request to my predecessor %d", node->pred_id); + + msg_error_t res = MSG_task_send_with_timeout(task_sent, mailbox, timeout); + + if (res != MSG_OK) { + XBT_DEBUG("Failed to send the 'Predecessor Alive' request (task %p) to %d", task_sent, node->pred_id); + task_free(task_sent); + }else{ + + // receive the answer + XBT_DEBUG("Sent 'Predecessor Alive' request (task %p) to %d, waiting for the answer on my mailbox '%s'", + task_sent, node->pred_id, req_data->answer_to); + + do { + if (node->comm_receive == NULL) { // FIXME simplify this + msg_task_t task_received = NULL; + node->comm_receive = MSG_task_irecv(&task_received, node->mailbox); + } + + res = MSG_comm_wait(node->comm_receive, timeout); + + if (res != MSG_OK) { + XBT_DEBUG("Failed to receive the answer to my 'Predecessor Alive' request (task %p): %d", + task_sent, (int)res); + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + node->pred_id = -1; + }else { + msg_task_t task_received = MSG_comm_get_task(node->comm_receive); + if (task_received != task_sent) { + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + handle_task(node, task_received); + }else{ + XBT_DEBUG("Received the answer to my 'Predecessor Alive' request (task %p) : my predecessor %d is alive", task_received, node->pred_id); + stop = 1; + MSG_comm_destroy(node->comm_receive); + node->comm_receive = NULL; + task_free(task_received); + } + } + } while (!stop); + } +} + +/** + * \brief Performs a find successor request to a random id. + * \param node the current node + */ +static void random_lookup(node_t node) +{ + + int id = 1337; + find_successor(node, id); + + /*** Random lookup disabled for tesh examples ***/ + /*if(node->stream == NULL) + node->stream = RngStream_CreateStream(""); + int random_index = RngStream_RandInt (node->stream, 0, nb_bits - 1); + int random_id = node->fingers[random_index].id; + XBT_DEBUG("Making a lookup request for id %d", random_id); + int res = find_successor(node, random_id); + XBT_DEBUG("The successor of node %d is %d", random_id, res);*/ + } /** @@ -963,28 +1024,50 @@ static void check_predecessor(node_t node) */ int main(int argc, char *argv[]) { + MSG_init(&argc, argv); if (argc < 3) { - printf("Usage: %s platform_file deployment_file\n", argv[0]); + printf("Usage: %s [-nb_bits=n] [-timeout=t] platform_file deployment_file\n", argv[0]); printf("example: %s ../msg_platform.xml chord.xml\n", argv[0]); exit(1); } - const char* platform_file = argv[1]; - const char* application_file = argv[2]; + char **options = &argv[1]; + while (!strncmp(options[0], "-", 1)) { + + int length = strlen("-nb_bits="); + if (!strncmp(options[0], "-nb_bits=", length) && strlen(options[0]) > length) { + nb_bits = atoi(options[0] + length); + XBT_DEBUG("Set nb_bits to %d", nb_bits); + } + else { + + length = strlen("-timeout="); + if (!strncmp(options[0], "-timeout=", length) && strlen(options[0]) > length) { + timeout = atoi(options[0] + length); + XBT_DEBUG("Set timeout to %d", timeout); + } + else { + xbt_die("Invalid chord option '%s'", options[0]); + } + } + options++; + } + + const char* platform_file = options[0]; + const char* application_file = options[1]; chord_initialize(); - MSG_global_init(&argc, argv); - MSG_set_channel_number(0); MSG_create_environment(platform_file); MSG_function_register("node", node); MSG_launch_application(application_file); - MSG_error_t res = MSG_main(); - INFO1("Simulation time: %g", MSG_get_clock()); + msg_error_t res = MSG_main(); + XBT_CRITICAL("Messages created: %ld", smx_total_comms); + XBT_INFO("Simulated time: %g", MSG_get_clock()); - MSG_clean(); + chord_exit(); if (res == MSG_OK) return 0;