1 /* Copyright (c) 2012, 2014-2016. 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 "dht-kademlia.h"
11 #include "simgrid/msg.h"
12 /** @addtogroup MSG_examples
13 * <b>kademlia/kademlia.c: Kademlia protocol</b>
14 * Implements the Kademlia protocol, using 32 bits identifiers.
16 XBT_LOG_NEW_DEFAULT_CATEGORY(msg_kademlia, "Messages specific for this msg example");
18 /* Main loop for the process */
19 static void main_loop(node_t node, double deadline)
21 double next_lookup_time = MSG_get_clock() + random_lookup_interval;
22 XBT_VERB("Main loop start");
24 while (MSG_get_clock() < deadline) {
25 if (node->receive_comm == NULL) {
26 node->task_received = NULL;
27 node->receive_comm = MSG_task_irecv(&node->task_received, node->mailbox);
30 if (node->receive_comm) {
31 if (MSG_comm_test(node->receive_comm) == 0) {
32 /* We search for a pseudo random node */
33 if (MSG_get_clock() >= next_lookup_time) {
35 next_lookup_time += random_lookup_interval;
37 //Didn't get a task: sleep for a while...
41 //There has been a transfer, we need to handle it !
42 msg_error_t status = MSG_comm_get_status(node->receive_comm);
43 MSG_comm_destroy(node->receive_comm);
44 node->receive_comm = NULL;
46 if (status == MSG_OK && node->task_received != NULL) {
47 xbt_assert((node->task_received != NULL), "We received an incorrect task");
48 handle_task(node, node->task_received);
50 xbt_die("A problem occurred. Comm ended in status '%d' and there was '%d' received task",
51 status, node->task_received == NULL ? 0 : 1);
55 //Didn't get a comm: sleep.
59 //Cleanup the receiving communication.
60 if (node->receive_comm != NULL) {
61 if (MSG_comm_test(node->receive_comm) != 0 && MSG_comm_get_status(node->receive_comm) == MSG_OK) {
62 task_free(MSG_comm_get_task(node->receive_comm));
64 MSG_comm_destroy(node->receive_comm);
68 /** @brief Node function
70 * @param the ID of the person I know in the system (or not)
71 * @param Time before I leave the system because I'm bored
73 static int node(int argc, char *argv[])
75 unsigned int join_sucess = 1;
77 xbt_assert(argc == 3 || argc == 4, "Wrong number of arguments");
78 /* Node initialization */
79 unsigned int id = strtoul(argv[1], NULL, 0);
80 node_t node = node_init(id);
83 XBT_INFO("Hi, I'm going to join the network with id %s", node->mailbox);
84 unsigned int id_known = strtoul(argv[2], NULL, 0);
85 join_sucess = join(node, id_known);
86 deadline = strtod(argv[3], NULL);
88 deadline = strtod(argv[2], NULL);
89 XBT_INFO("Hi, I'm going to create the network with id %s", node->mailbox);
90 node_routing_table_update(node, node->id);
93 XBT_VERB("Ok, I'm joining the network with id %s", node->mailbox);
94 //We start the main loop
95 main_loop(node, deadline);
97 XBT_INFO("I couldn't join the network :(");
99 XBT_DEBUG("I'm leaving the network");
100 XBT_INFO("%d/%d FIND_NODE have succeeded", node->find_node_success, node->find_node_success + node->find_node_failed);
107 * @brief Tries to join the network
108 * @param node node data
109 * @param id_known id of the node I know in the network.
111 unsigned int join(node_t node, unsigned int id_known)
115 unsigned int trial = 0;
117 unsigned int answer_got = 0;
119 /* Add the guy we know to our routing table and ourselves. */
120 node_routing_table_update(node, node->id);
121 node_routing_table_update(node, id_known);
123 /* First step: Send a "FIND_NODE" request to the node we know */
124 send_find_node(node, id_known, node->id);
126 if (node->receive_comm == NULL) {
127 node->task_received = NULL;
128 node->receive_comm = MSG_task_irecv(&node->task_received, node->mailbox);
130 if (node->receive_comm) {
131 if (MSG_comm_test(node->receive_comm)) {
132 status = MSG_comm_get_status(node->receive_comm);
133 MSG_comm_destroy(node->receive_comm);
134 node->receive_comm = NULL;
135 if (status == MSG_OK) {
136 XBT_DEBUG("Received an answer from the node I know.");
138 //retrieve the node list and ping them.
139 task_data_t data = MSG_task_get_data(node->task_received);
140 if (data && data->type == TASK_FIND_NODE_ANSWER) {
141 node_contact_t contact;
142 node_list = data->answer;
143 xbt_dynar_foreach(node_list->nodes, i, contact) {
144 node_routing_table_update(node, contact->id);
146 task_free(node->task_received);
148 handle_task(node, node->task_received);
154 MSG_process_sleep(1);
157 MSG_process_sleep(1);
159 } while (answer_got == 0 && trial < max_join_trials);
160 /* Second step: Send a FIND_NODE to a a random node in buckets */
161 unsigned int bucket_id = routing_table_find_bucket(node->table, id_known)->id;
162 for (i = 0; ((bucket_id - i) > 0 || (bucket_id + i) <= identifier_size) && i < JOIN_BUCKETS_QUERIES; i++) {
163 if (bucket_id - i > 0) {
164 unsigned int id_in_bucket = get_id_in_prefix(node->id, bucket_id - i);
165 find_node(node, id_in_bucket, 0);
167 if (bucket_id + i <= identifier_size) {
168 unsigned int id_in_bucket = get_id_in_prefix(node->id, bucket_id + i);
169 find_node(node, id_in_bucket, 0);
175 /** @brief Send a request to find a node in the node routing table.
176 * @param node our node data
177 * @param id_to_find the id of the node we are trying to find
179 unsigned int find_node(node_t node, unsigned int id_to_find, unsigned int count_in_stats)
182 unsigned int queries;
183 unsigned int answers;
184 unsigned int destination_found = 0;
185 unsigned int nodes_added = 0;
186 double time_beginreceive;
188 double global_timeout = MSG_get_clock() + find_node_global_timeout;
189 unsigned int steps = 0;
191 xbt_assert((id_to_find >= 0), "Id supplied incorrect");
193 /* First we build a list of who we already know */
194 answer_t node_list = node_find_closest(node, id_to_find);
195 xbt_assert((node_list != NULL), "node_list incorrect");
197 XBT_DEBUG("Doing a FIND_NODE on %08x", id_to_find);
201 /* Ask the nodes on our list if they have information about the node we are trying to find */
204 queries = send_find_node_to_best(node, node_list);
206 timeout = MSG_get_clock() + find_node_timeout;
208 time_beginreceive = MSG_get_clock();
210 if (node->receive_comm == NULL) {
211 node->task_received = NULL;
212 node->receive_comm = MSG_task_irecv(&node->task_received, node->mailbox);
214 if (node->receive_comm) {
215 if (MSG_comm_test(node->receive_comm)) {
216 status = MSG_comm_get_status(node->receive_comm);
217 MSG_comm_destroy(node->receive_comm);
218 node->receive_comm = NULL;
219 if (status == MSG_OK) {
220 xbt_assert((node->task_received != NULL), "Invalid task received");
221 //Figure out if we received an answer or something else
222 task_data_t data = MSG_task_get_data(node->task_received);
224 //Check if what we have received is what we are looking for.
225 if (data && data->type == TASK_FIND_NODE_ANSWER && data->answer->destination_id == id_to_find) {
227 node_routing_table_update(node, data->sender_id);
228 node_contact_t contact;
229 xbt_dynar_foreach(node_list->nodes, i, contact) {
230 node_routing_table_update(node, contact->id);
234 nodes_added = answer_merge(node_list, data->answer);
235 XBT_DEBUG("Received an answer from %s (%s) with %ld nodes on it",
236 data->answer_to, data->issuer_host_name, xbt_dynar_length(data->answer->nodes));
238 task_free(node->task_received);
240 handle_task(node, node->task_received);
241 //Update the timeout if we didn't have our answer
242 timeout += MSG_get_clock() - time_beginreceive;
243 time_beginreceive = MSG_get_clock();
247 MSG_process_sleep(1);
250 MSG_process_sleep(1);
252 } while (MSG_get_clock() < timeout && answers < queries);
253 destination_found = answer_destination_found(node_list);
254 } while (!destination_found && (nodes_added > 0 || answers == 0) && MSG_get_clock() < global_timeout
255 && steps < MAX_STEPS);
256 if (destination_found) {
258 node->find_node_success++;
260 XBT_VERB("FIND_NODE on %08x success in %d steps", id_to_find, steps);
261 node_routing_table_update(node, id_to_find);
263 if (count_in_stats) {
264 node->find_node_failed++;
265 XBT_VERB("%08x not found in %d steps", id_to_find, steps);
268 answer_free(node_list);
269 return destination_found;
272 /** @brief Pings a node in the system to see if it is online.
273 * @param node Our node data
274 * @param id_to_ping the id of a node we want to see if it is online.
275 * @return if the ping succeded or not.
277 unsigned int ping(node_t node, unsigned int id_to_ping)
279 char mailbox[MAILBOX_NAME_SIZE];
280 snprintf(mailbox,MAILBOX_NAME_SIZE, "%d", id_to_ping);
282 unsigned int destination_found = 0;
283 double timeout = MSG_get_clock() + ping_timeout;
285 msg_task_t ping_task = task_new_ping(node->id, node->mailbox, MSG_host_get_name(MSG_host_self()));
286 msg_task_t task_received = NULL;
288 XBT_VERB("PING %08x", id_to_ping);
290 //Check that we aren't trying to ping ourselves
291 if (id_to_ping == node->id) {
295 /* Sending the ping task */
296 MSG_task_dsend(ping_task, mailbox, task_free_v);
298 task_received = NULL;
300 MSG_task_receive_with_timeout(&task_received, node->mailbox, ping_timeout);
301 if (status == MSG_OK && task_received) {
302 //Checking if it's what we are waiting for or not.
303 task_data_t data = MSG_task_get_data(task_received);
304 if (data && data->type == TASK_PING_ANSWER && id_to_ping == data->sender_id) {
305 XBT_VERB("Ping to %s succeeded", mailbox);
306 node_routing_table_update(node, data->sender_id);
307 destination_found = 1;
308 task_free(task_received);
310 //If it's not our answer, we answer the query anyway.
311 handle_task(node, task_received);
314 } while (destination_found == 0 && MSG_get_clock() < timeout);
316 if (MSG_get_clock() >= timeout) {
317 XBT_DEBUG("Ping to %s has timeout.", mailbox);
320 if (destination_found == -1) {
321 XBT_DEBUG("It seems that %s is offline...", mailbox);
327 /** @brief Does a pseudo-random lookup for someone in the system
328 * @param node caller node data
330 void random_lookup(node_t node)
332 unsigned int id_to_look = RANDOM_LOOKUP_NODE; //Totally random.
333 /* TODO: Use some pseudorandom generator like RngStream. */
334 XBT_DEBUG("I'm doing a random lookup");
335 find_node(node, id_to_look, 1);
338 /** @brief Send a "FIND_NODE" to a node
339 * @param node sender node data
340 * @param id node we are querying
341 * @param destination node we are trying to find.
343 void send_find_node(node_t node, unsigned int id, unsigned int destination)
345 char mailbox[MAILBOX_NAME_SIZE];
346 /* Gets the mailbox to send to */
347 get_node_mailbox(id, mailbox);
349 msg_task_t task = task_new_find_node(node->id, destination, node->mailbox, MSG_host_get_name(MSG_host_self()));
351 xbt_assert((task != NULL), "Trying to send a NULL task.");
352 MSG_task_dsend(task, mailbox, task_free_v);
353 XBT_VERB("Asking %s for its closest nodes", mailbox);
357 * Sends to the best "kademlia_alpha" nodes in the "node_list" array a "FIND_NODE" request, to ask them for their best nodes
359 unsigned int send_find_node_to_best(node_t node, answer_t node_list)
363 unsigned int destination = node_list->destination_id;
364 node_contact_t node_to_query;
365 while (j < kademlia_alpha && i < node_list->size) {
366 /* We need to have at most "kademlia_alpha" requests each time, according to the protocol */
367 /* Gets the node we want to send the query to */
368 node_to_query = xbt_dynar_get_as(node_list->nodes, i, node_contact_t);
369 if (node_to_query->id != node->id) { /* No need to query ourselves */
370 send_find_node(node, node_to_query->id, destination);
378 /** @brief Handles an incoming received task */
379 void handle_task(node_t node, msg_task_t task)
381 task_data_t data = MSG_task_get_data(task);
383 //Adding/updating the guy to our routing table
384 node_routing_table_update(node, data->sender_id);
385 switch (data->type) {
387 handle_find_node(node, data);
389 case TASK_FIND_NODE_ANSWER:
390 XBT_DEBUG("Received a wrong answer for a FIND_NODE");
393 handle_ping(node, data);
402 /** @brief Handles the answer to an incoming "find_node" task */
403 void handle_find_node(node_t node, task_data_t data)
405 XBT_VERB("Received a FIND_NODE from %s (%s), he's trying to find %08x",
406 data->answer_to, data->issuer_host_name, data->destination_id);
407 //Building the answer to the request
408 answer_t answer = node_find_closest(node, data->destination_id);
409 //Building the task to send
410 msg_task_t task = task_new_find_node_answer(node->id, data->destination_id, answer, node->mailbox,
411 MSG_host_get_name(MSG_host_self()));
413 MSG_task_dsend(task, data->answer_to, task_free_v);
416 /** @brief handles the answer to a ping */
417 void handle_ping(node_t node, task_data_t data)
419 XBT_VERB("Received a PING request from %s (%s)", data->answer_to, data->issuer_host_name);
420 //Building the answer to the request
421 msg_task_t task = task_new_ping_answer(node->id, data->answer_to, MSG_host_get_name(MSG_host_self()));
423 MSG_task_dsend(task, data->answer_to, task_free_v);
426 /** @brief Main function */
427 int main(int argc, char *argv[])
429 MSG_init(&argc, argv);
431 /* Check the arguments */
432 xbt_assert(argc > 2, "Usage: %s platform_file deployment_file\n\tExample: %s msg_platform.xml msg_deployment.xml\n",
435 const char *platform_file = argv[1];
436 const char *deployment_file = argv[2];
438 MSG_create_environment(platform_file);
439 MSG_function_register("node", node);
440 MSG_launch_application(deployment_file);
442 msg_error_t res = MSG_main();
444 XBT_INFO("Simulated time: %g", MSG_get_clock());
446 return res != MSG_OK;