-/* 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
#include "msg/msg.h"
#include "xbt/log.h"
#include "xbt/asserts.h"
-#include "mc/modelchecker.h"
-#include "mc/mc.h"
+#include "simgrid/modelchecker.h"
+#include <xbt/RngStream.h>
+
+/** @addtogroup MSG_examples
+ *
+ * - <b>chord/chord.c: Classical Chord P2P protocol</b>
+ * 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");
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[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[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)
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()
msg_comm_t comm_receive; // current communication to receive
- xbt_dynar_t comms; // current communications being sent
double last_change_date; // last time I changed a finger or my predecessor
+ RngStream stream; //RngStream for
} s_node_t, *node_t;
/**
TASK_GET_PREDECESSOR_ANSWER,
TASK_NOTIFY,
TASK_SUCCESSOR_LEAVING,
- TASK_PREDECESSOR_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 task_data {
+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)
// 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 void get_mailbox(int host_id, char* mailbox);
-static void task_data_destroy(task_data_t task_data);
+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, m_task_t task);
+static void handle_task(node_t node, msg_task_t task);
// Chord core
static void create(node_t node);
static void fix_fingers(node_t node);
static void check_predecessor(node_t node);
static void random_lookup(node_t);
-static void quit_notify(node_t node, int to);
+static void quit_notify(node_t node);
/**
* \brief Global initialization of the Chord simulation.
pow = pow << 1;
}
nb_keys = pow;
- DEBUG1("Sets nb_keys to %d", nb_keys);
+ XBT_DEBUG("Sets nb_keys to %d", nb_keys);
+}
+
+static void chord_exit(void)
+{
+ xbt_free(powers2);
}
/**
}
/**
- * \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 nb_keys - 1).
* 1 belongs to [62, 3]
}
/**
- * \brief Frees the memory used by some task data.
- * \param task_data the task data to destroy
+ * \brief Frees the memory used by a task.
+ * \param task the MSG task to destroy
*/
-static void task_data_destroy(task_data_t task_data)
+static void task_free(void* task)
{
- xbt_free(task_data);
+ // 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);
+ }
}
/**
{
if (XBT_LOG_ISENABLED(msg_chord, xbt_log_priority_verbose)) {
int i;
- VERB0("My finger table:");
- VERB0("Start | Succ ");
+ XBT_VERB("My finger table:");
+ XBT_VERB("Start | Succ ");
for (i = 0; i < nb_bits; i++) {
- VERB2(" %3d | %3d ", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
+ XBT_VERB(" %3d | %3d ", (node->id + powers2[i]) % nb_keys, node->fingers[i].id);
}
- VERB1("Predecessor: %d", node->pred_id);
+ XBT_VERB("Predecessor: %d", node->pred_id);
}
}
node->fingers[finger_index].id = id;
get_mailbox(id, node->fingers[finger_index].mailbox);
node->last_change_date = MSG_get_clock();
- DEBUG2("My new finger #%d is %d", finger_index, id);
+ XBT_DEBUG("My new finger #%d is %d", finger_index, id);
}
}
}
node->last_change_date = MSG_get_clock();
- DEBUG1("My new predecessor is %d", predecessor_id);
+ XBT_DEBUG("My new predecessor is %d", 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_received = NULL;
- msg_comm_t comm_send = NULL;
+ msg_task_t task_received = NULL;
int i;
- int index;
int join_success = 0;
double deadline;
double next_stabilize_date = init_time + periodic_stabilize_delay;
double next_check_predecessor_date = init_time + periodic_check_predecessor_delay;
double next_lookup_date = init_time + periodic_lookup_delay;
- xbt_assert0(argc == 3 || argc == 5, "Wrong number of arguments for this node");
+ #ifdef HAVE_MC
+ int listen = 0;
+ int no_op = 0;
+ int sub_protocol = 0;
+ #endif
+
+ 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]);
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;
/*
// sleep before starting
- DEBUG1("Let's sleep during %f", sleep_time);
+ XBT_DEBUG("Let's sleep during %f", sleep_time);
MSG_process_sleep(sleep_time);
*/
- DEBUG0("Hey! Let's join the system.");
+ XBT_DEBUG("Hey! Let's join the system.");
join_success = join(&node, known_id);
}
if (join_success) {
while (MSG_get_clock() < init_time + deadline
-// && MSG_get_clock() < node.last_change_date + 1000
- && MSG_get_clock() < max_simulation_time) {
+// && MSG_get_clock() < node.last_change_date + 1000
+ && MSG_get_clock() < max_simulation_time) {
if (node.comm_receive == NULL) {
task_received = NULL;
// FIXME: do not make MSG_task_irecv() calls from several functions
}
+ //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) {
+ }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) {
+ }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 {
+ }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 {
- // a transfer has occured
+#endif
+
+ } else {
+ // a transfer has occurred
- MSG_error_t status = MSG_comm_get_status(node.comm_receive);
+ msg_error_t status = MSG_comm_get_status(node.comm_receive);
if (status != MSG_OK) {
- DEBUG0("Failed to receive a task. Nevermind.");
+ XBT_DEBUG("Failed to receive a task. Nevermind.");
+ MSG_comm_destroy(node.comm_receive);
node.comm_receive = NULL;
}
else {
handle_task(&node, task_received);
}
}
-
- // see if some communications are finished
- while ((index = MSG_comm_testany(node.comms)) != -1) {
- comm_send = xbt_dynar_get_as(node.comms, index, msg_comm_t);
- MSG_error_t status = MSG_comm_get_status(comm_send);
- xbt_dynar_remove_at(node.comms, index, &comm_send);
- DEBUG3("Communication %p is finished with status %d, dynar size is now %lu",
- comm_send, status, xbt_dynar_length(node.comms));
- m_task_t task = MSG_comm_get_task(comm_send);
- MSG_comm_destroy(comm_send);
- if (status != MSG_OK) {
- task_data_destroy(MSG_task_get_data(task));
- MSG_task_destroy(task);
- }
- }
}
- // clean unfinished comms sent
- /* unsigned int cursor;
- xbt_dynar_foreach(node.comms, cursor, comm_send) {
- m_task_t task = MSG_comm_get_task(comm_send);
- MSG_task_cancel(task);
- task_data_destroy(MSG_task_get_data(task));
- MSG_task_destroy(task);
- MSG_comm_destroy(comm_send);
- // FIXME: the task is actually not destroyed because MSG thinks that the other side (whose process is dead) is still using it
- }*/
+ if (node.comm_receive) {
+ MSG_comm_destroy(node.comm_receive);
+ node.comm_receive = NULL;
+ }
// leave the ring
leave(&node);
}
// stop the simulation
- xbt_dynar_free(&node.comms);
xbt_free(node.fingers);
return 0;
}
* \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, m_task_t task) {
+static void handle_task(node_t node, msg_task_t task) {
- DEBUG1("Handling task %p", task);
- msg_comm_t comm = NULL;
+ 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;
switch (type) {
- case TASK_FIND_SUCCESSOR:
- DEBUG2("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;
- DEBUG4("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);
- 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);
- DEBUG2("Forwarding the 'Find Successor' request for id %d to my closest preceding finger %d",
- task_data->request_id, closest);
- get_mailbox(closest, mailbox);
- comm = MSG_task_isend(task, mailbox);
- xbt_dynar_push(node->comms, &comm);
- }
- break;
-
- case TASK_GET_PREDECESSOR:
- DEBUG1("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;
- DEBUG3("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);
- comm = MSG_task_isend(task, task_data->answer_to);
- xbt_dynar_push(node->comms, &comm);
- break;
-
- case TASK_NOTIFY:
- // someone is telling me that he may be my new predecessor
- DEBUG1("Receiving a 'Notify' request from %s", task_data->issuer_host_name);
- notify(node, task_data->request_id);
- task_data_destroy(task_data);
- MSG_task_destroy(task);
- break;
-
- case TASK_PREDECESSOR_LEAVING:
- // my predecessor is about to quit
- DEBUG1("Receiving a 'Predecessor Leaving' message from %s", task_data->issuer_host_name);
- // modify my predecessor
- set_predecessor(node, task_data->request_id);
- task_data_destroy(task_data);
- MSG_task_destroy(task);
- /*TODO :
+ 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
- DEBUG1("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_data_destroy(task_data);
- MSG_task_destroy(task);
- /* TODO
- >> notify my new successor
- >> update my table & predecessors table */
- break;
-
- case TASK_FIND_SUCCESSOR_ANSWER:
- case TASK_GET_PREDECESSOR_ANSWER:
- DEBUG2("Ignoring unexpected task of type %d (%p)", type, task);
- task_data_destroy(task_data);
- MSG_task_destroy(task);
- break;
+ */
+ 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;
+
}
}
*/
static void create(node_t node)
{
- DEBUG0("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);
}
*/
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 i;
+ for (i = 0; i < nb_bits; i++) {
+ set_finger(node, i, known_id);
+ }
+ */
+
int successor_id = remote_find_successor(node, known_id, node->id);
if (successor_id == -1) {
- INFO0("Cannot join the ring.");
+ XBT_INFO("Cannot join the ring.");
}
else {
set_finger(node, 0, successor_id);
*/
static void leave(node_t node)
{
- DEBUG0("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)
{
- /* TODO
- 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());
- req_data->answer_to = NULL;
- 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);
- req_data->type = TASK_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);
- req_data->type = TASK_SUCCESSOR_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(NULL, COMP_SIZE, COMM_SIZE, req_data);
- //char* mailbox = get_mailbox(to_mailbox);
- msg_comm_t comm = MSG_task_isend(task, to_mailbox);
- xbt_dynar_push(node->comms, &comm);
- */
+
+ //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);
+ }
+
}
/**
req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
// send a "Find Successor" request to ask_to_id
- m_task_t task_sent = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
- DEBUG3("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);
+ 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) {
- DEBUG3("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
+ XBT_DEBUG("Failed to send the 'Find Successor' request (task %p) to %d for id %d",
task_sent, ask_to, id);
- MSG_task_destroy(task_sent);
- task_data_destroy(req_data);
+ task_free(task_sent);
}
else {
// receive the answer
- DEBUG3("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
+ XBT_DEBUG("Sent a 'Find Successor' request (task %p) to %d for key %d, waiting for the answer",
task_sent, ask_to, id);
do {
if (node->comm_receive == NULL) {
- m_task_t task_received = NULL;
+ 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) {
- DEBUG2("Failed to receive the answer to my 'Find Successor' request (task %p): %d",
- task_sent, res);
+ 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;
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
}
else {
- m_task_t task_received = MSG_comm_get_task(node->comm_receive);
- DEBUG1("Received a task (%p)", task_received);
+ 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);
- if (MC_IS_ENABLED) {
- MC_assert(task_received == task_sent);
- }
+ // 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;
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
handle_task(node, task_received);
}
else {
// this is our answer
- DEBUG4("Received the answer to my 'Find Successor' request for id %d (task %p): the successor of key %d is %d",
+ 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;
- MSG_task_destroy(task_received);
- task_data_destroy(req_data);
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
+ task_free(task_received);
}
}
} while (!stop);
req_data->issuer_host_name = MSG_host_get_name(MSG_host_self());
// send a "Get Predecessor" request to ask_to_id
- DEBUG1("Sending a 'Get Predecessor' request to %d", ask_to);
- m_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);
+ 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) {
- DEBUG2("Failed to send the 'Get Predecessor' request (task %p) to %d",
+ XBT_DEBUG("Failed to send the 'Get Predecessor' request (task %p) to %d",
task_sent, ask_to);
- MSG_task_destroy(task_sent);
- task_data_destroy(req_data);
+ task_free(task_sent);
}
else {
// receive the answer
- DEBUG3("Sent 'Get Predecessor' request (task %p) to %d, waiting for the answer on my mailbox '%s'",
+ 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);
do {
if (node->comm_receive == NULL) { // FIXME simplify this
- m_task_t task_received = NULL;
+ 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) {
- DEBUG2("Failed to receive the answer to my 'Get Predecessor' request (task %p): %d",
- task_sent, res);
+ 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;
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
}
else {
- m_task_t task_received = MSG_comm_get_task(node->comm_receive);
+ 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_ENABLED) {
- MC_assert(task_received == task_sent);
- }
+ /*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;
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
handle_task(node, task_received);
}
else {
- DEBUG3("Received the answer to my 'Get Predecessor' request (task %p): the predecessor of node %d is %d",
+ 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;
- MSG_task_destroy(task_received);
- task_data_destroy(req_data);
+ MSG_comm_destroy(node->comm_receive);
+ node->comm_receive = NULL;
+ task_free(task_received);
}
}
} while (!stop);
*/
static void stabilize(node_t node)
{
- DEBUG0("Stabilizing node");
+ XBT_DEBUG("Stabilizing node");
// get the predecessor of my immediate successor
int candidate_id;
print_finger_table(node);
}
else {
- DEBUG1("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);
}
}
*/
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->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
- m_task_t task = MSG_task_create(NULL, COMP_SIZE, COMM_SIZE, req_data);
- DEBUG2("Sending a 'Notify' request (task %p) to %d", task, notify_id);
- char mailbox[MAILBOX_NAME_SIZE];
- get_mailbox(notify_id, mailbox);
- msg_comm_t comm = MSG_task_isend(task, mailbox);
- xbt_dynar_push(node->comms, &comm);
-}
+ 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.
*/
static void fix_fingers(node_t node) {
- DEBUG0("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) {
*/
static void check_predecessor(node_t node)
{
- DEBUG0("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);
+ }
}
/**
*/
static void random_lookup(node_t node)
{
- int id = 1337; // TODO pick a pseudorandom id
- DEBUG1("Making a lookup request for id %d", id);
+
+ 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);*/
+
}
/**
*/
int main(int argc, char *argv[])
{
+ MSG_init(&argc, argv);
if (argc < 3) {
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);
}
- MSG_global_init(&argc, argv);
-
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);
- DEBUG1("Set nb_bits to %d", nb_bits);
+ 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);
- DEBUG1("Set timeout to %d", timeout);
+ timeout = atoi(options[0] + length);
+ XBT_DEBUG("Set timeout to %d", timeout);
}
else {
- xbt_assert1(0, "Invalid chord option '%s'", options[0]);
+ xbt_die("Invalid chord option '%s'", options[0]);
}
}
options++;
chord_initialize();
- 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;
